A water content early warning method for near infrared online analysis of distillate oil products in atmospheric and vacuum unit

By using principal component analysis to monitor the water content and spectral distortion at the bottom of the optical cell, the problem of false alarms in near-infrared spectrometers caused by increased water content in oil products after the modification of atmospheric and vacuum distillation units in refining and chemical enterprises was solved, thus achieving reliable online analysis and production stability.

CN114839155BActive Publication Date: 2024-11-22NANJING RICHISLAND INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210082359.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-24
Publication Date
2024-11-22
Estimated Expiration
2042-01-24

AI Technical Summary

Technical Problem

After the atmospheric and vacuum distillation unit is upgraded, the increased water content in the oil products leads to water droplet deposition at the bottom of the near-infrared spectrometer's photocell, affecting the accuracy of online analysis results. It also makes it difficult to distinguish between the effects of water and spectral changes caused by changes in operating conditions, resulting in false alarms and increased maintenance burden.

Method used

A model was established using principal component analysis to monitor the water content and spectral distortion at the bottom of the photocell. By monitoring spectral data through modeling, the influence of moisture and changes in operating conditions were distinguished, enabling early warning and automatic identification of operating condition switching, thereby reducing false alarms.

Benefits of technology

It enables reliable online analysis of oil properties, reduces false alarm rates, alleviates the maintenance burden on refining and chemical enterprises, and improves the accuracy of analysis results and the stability of production.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114839155B_ABST
    Figure CN114839155B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for early warning of water content in near-infrared online analysis of distillate oil products in a constant-pressure reduction unit. The method uses principal component analysis to establish a dual model for data under normal operating conditions, wherein the first model monitors and warns the water content in the online analysis of oil product properties, and the second model monitors and alarms the spectral spectrum distortion caused by water. The method has good tolerance to near-infrared spectral distortion caused by water, so as to reduce the maintenance burden of refining enterprises. Finally, the method can automatically identify the operating condition changes caused by the change of oil product properties, reduce the false alarm rate of online analysis, and reduce the labor intensity of online analysis maintenance personnel.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of near-infrared spectroscopy online analysis, and in particular to a water content early warning method in near-infrared spectroscopy online analysis of side-line oil products in a constant pressure reduction unit of a refinery. Background Art

[0002] In refining and chemical enterprises, in order to increase production capacity, existing atmospheric and vacuum distillation units are sometimes modified (such as increasing the height and the number of plates). This requires the introduction of more water vapor during production operations to drive the light components to move upward, thereby increasing the water content in the light component oil products.

[0003] The temperature during the near-infrared online analysis of oil products is much lower than that of the atmospheric and vacuum device, which causes the water content in the oil products to condense into water droplets. When the oil products continuously pass through the optical pool of the near-infrared spectrometer for a long time, it is inevitable that the water droplets will settle to the bottom of the optical pool, causing the near-infrared light emitted from the bottom of the optical pool to be absorbed or refracted by water, changing the baseline position of the near-infrared spectrum and the amplitude of each absorption peak, which will affect the accuracy of online analysis of oil properties. Therefore, it is necessary to monitor and warn the influence of water in the online analysis of oil properties in order to ensure timely maintenance and ensure accurate and reliable online analysis results.

[0004] In fact, low levels of water have limited impact on the spectrum. If an alarm is triggered once water is detected, it will inevitably increase the maintenance burden of the refinery. Considering the continuity and stability of the production process, a certain degree of spectral distortion can be tolerated to cause deviations in detection properties. However, as the online analysis process proceeds, more and more water will accumulate at the bottom of the optical pool, eventually causing the spectral distortion to exceed the tolerance range, resulting in a large deviation in the online analysis results.

[0005] It should be pointed out that the oil composition in the online analysis process will also change with the change of raw materials and production process, that is, the production conditions will change. How to distinguish the spectral changes caused by the influence of moisture and the change of working conditions is also a key to reduce the false alarm of water content warning in online analysis. Summary of the invention

[0006] In view of the above problems, the present invention discloses a water content early warning method in near-infrared online analysis of oil products. The method uses principal component analysis to establish a model, monitors the water content at the bottom of the online analysis light pool and the influence of water on the near-infrared spectrum, identifies the change of working conditions and switches.

[0007] The present invention is divided into two parts: offline modeling and online monitoring. The modeling process includes early warning modeling of water content at the bottom of the optical pool and spectral distortion alarm modeling. The specific steps are as follows:

[0008] (1) Sampling multiple groups of online analysis of near-infrared spectral data and their properties;

[0009] (2) Analyze the near-infrared spectrum and select a smooth and water-containing sample with significant baseline drift. 1 ~a 2 cm -1 Wavenumber band spectral data X 1 For early warning of water content in oil products, select the oil component with significant absorption peak b 1 ~b 2 cm -1 Wavenumber band spectral data X 2 Used for property detection deviation alarm;

[0010] (3) Select X 1 The principal component analysis method is used to establish the fault warning model M based on the samples with normal working conditions. 1 , and obtain the control limit SPE α1 and

[0011] (4) Compare the properties calculated by near infrared spectroscopy data modeling with those obtained by chemical analysis, and use the deviation percentage β as the threshold to set X 2 The spectra in are divided into two categories, one of which is the spectral data Z that can tolerate the deviation of properties. 1 , and the other type is spectral data Z with intolerable property deviation 2 ;

[0012] (5) Using spectral data Z 1 The principal component analysis method is used to establish the fault warning model M 2 , and obtain the control limit SPE α2 and The model M 2 Added to the working condition library;

[0013] (6) Simultaneous use of model M 1 and Model M 2 Online monitoring of oil properties;

[0014] (7) According to model M 1 Monitoring spectral data statistics SPE 1 、T 1 2 And model M 2 Monitoring spectral data statistics SPE 2 , Perform different actions:

[0015] If the statistic SPE of the monitored spectral data 1 、T 1 2 and SPE 2 , If the oil does not exceed the limit, the oil to be tested does not contain water, and the process goes to step (6);

[0016] If SPE 1 、T 1 2 Exceeding the limit and SPE 2 , If the limit is not exceeded, the warning state is entered. At this time, the oil to be tested contains water, but the impact on the near-infrared oil property detection is within the tolerable range, and the process goes to step (6);

[0017] If SPE 1 、T 1 2 and SPE 2 , If both exceed the limit, the alarm state is entered. At this time, the influence of the water component of the oil product to be tested on the near-infrared oil product property detection exceeds the tolerance range, the online monitoring device is terminated, the water and impurities accumulated in the optical pool of the online spectrometer are cleaned, the online monitoring device is restarted, and the process goes to step (6);

[0018] If SPE 1 、T 1 2 Not exceeded but SPE 2 , If the limit is exceeded, the oil enters the operating condition transition state. At this time, the oil does not contain water, but the oil composition has changed, resulting in the oil spectrum X 2 The absorption peak changes significantly, and the spectral data X after the working conditions change is calculated. 2 When the consistency meets the requirements, it means that the working condition tends to be stable. Determine whether the working condition already exists in the working condition library. If the working condition does not appear in the library, go to step (4).

[0019] Beneficial effects:

[0020] The present invention discloses a method for early warning of water content in near-infrared online analysis of oil products. The method monitors and warns the influence of water content in online analysis of oil properties, so as to perform predictive maintenance and ensure the reliability of online analysis results. In addition, the method has good tolerance to near-infrared spectrum distortion caused by water, so as to reduce the maintenance burden of refining enterprises. Finally, the method can automatically identify the changes in operating conditions caused by changes in oil properties, reduce the false alarm rate of online analysis, and reduce the labor intensity of online analysis maintenance personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of online analysis of water content monitoring and early warning by the method of the present invention;

[0022] Figure 2 This is the spectrum of the total distillation naphtha of a certain refinery on July 12 in the embodiment;

[0023] Figure 3It is the monitoring result when the water content exceeds the threshold during the online analysis process;

[0024] Figure 4 It is the monitoring result when the spectral distortion exceeds the threshold during the online analysis process;

[0025] Figure 5 It is a diagram of the spectrum consistency judgment process when the working conditions change.

[0026] Specific implementation process

[0027] The following is a detailed calculation process and a specific operation flow to illustrate the implementation effect of the method in the online analysis process in combination with the accompanying drawings and embodiments. This implementation case is implemented based on the technical solution of the present invention, but the protection scope of the present invention is not limited to the following embodiments.

[0028] This embodiment is an online analysis of the properties of the atmospheric tower top naphtha of a certain refinery. In order to increase production capacity, the refinery transformed the atmospheric and vacuum unit in 2015, increasing the height and number of plates of the atmospheric tower. Therefore, it is necessary to introduce more water vapor to drive the light components to move upward, resulting in more water in the atmospheric tower top naphtha. During the online analysis process, the water in the naphtha easily condenses into water droplets and deposits at the bottom of the optical pool of the online analyzer, resulting in spectrum distortion and deviation of the online monitoring properties.

[0029] In conjunction with this embodiment, the specific implementation steps of this method are as follows Figure 1 As shown, the details are as follows:

[0030] (1) The wave number range of the top naphtha collected from the refinery in July was 4500-10000 cm -1 Near-infrared spectra and corresponding data on the distillation properties of oil products.

[0031] (2) Figure 2 As shown in the figure, the water content of the distilled naphtha of the enterprise on July 12 changed, with 4705~5550cm -1 The wavenumber spectrum is smooth and the water content causes a significant baseline drift, 5639-5921cm -1 The absorption peaks of oil components in the wave number band spectrum are significant. Therefore, the former band is suitable for monitoring the moisture content of oil products, and the latter band is suitable for monitoring the distillation range property deviation;

[0032] (3) According to the process data, the near infrared spectrometer had just been dewatered on July 3, and the water content in the optical pool was low, which was suitable for water content modeling analysis. Therefore, the 4705-5550cm -1 The principal component analysis method was used to establish the water content early warning model M based on the wave number band spectrum. 1 , and obtain the control limit SPE α1 and The specific process is as follows:

[0033] ①To X 1 Perform column standardization and find X 1 The covariance matrix S of is taken and the eigendecomposition is performed. The results are as follows:

[0034]

[0035] [λ 1 ,λ 2 ,…,λ 220 ]=[219.788 0.152 … 0] 1×220

[0036]

[0037] Where P is the loading matrix, and the eigenvalues ​​in the diagonal matrix satisfy λ 1 ≥λ 2 ≥…≥λ n , according to the requirements of the principal component contribution, select the first principal component, that is, k = 1, and the principal component load P = p 1 .

[0038] ② Calculate the control line SPE α1 and

[0039] α=0.05

[0040]

[0041]

[0042] c α =1.6449

[0043] F(k,nk,α)=F(1,219,0.05)=0.0028

[0044]

[0045]

[0046] ③Select the 4705~5550cm of the online spectral data on July 12 -1 The wave number band spectrum is used to calculate the monitoring indicators SPE and T according to the following formula 2 :

[0047] SPE=||(I-PP T )x|| 2

[0048] T 2 =x TPΛ -1 P T x

[0049] Monitoring results such as Figure 3 As shown, when SPE or T 2 If the control limit is exceeded, the bottom of the optical pool contains water, which shows that this method successfully detects the water content of the oil;

[0050] (4) Compare the properties calculated by near infrared spectroscopy data modeling with those obtained by chemical analysis, and use the deviation percentage β = 5% as the threshold to set X 2 The spectra in are divided into two categories, one of which is the spectral data Z that can tolerate the deviation of properties. 1 , and the other type is spectral data Z with intolerable property deviation 2 ;

[0051] (5) By comparing the online analysis with the traditional test results, the absorption peaks of the oil components at 5639-5921 cm were extracted from the online analysis spectrum. -1 Wave number band spectral data Z 1 Used for spectrum distortion alarm. 1 The principal component analysis method was used to establish the spectral distortion alarm model M 2 , and obtain the control limit SPE α2 and The specific steps are the same as (3), and the control limit SPE is obtained α2 =0.0327, Take the relevant spectrum of some online spectrum data on July 3 for monitoring, the monitoring results are as follows Figure 4 As shown, it can be seen that distortion occurs from the 22nd sample and exceeds the threshold;

[0052] (6) Simultaneous use of model M 1 and Model M 2 Online monitoring of oil properties;

[0053] (7) By monitoring the relevant spectra of the online spectral data on July 25, it was found that SPE appeared in the spectrum from 10:00 to 12:00. 1 and T 1 2 Not exceeded, but SPE 2 and Exceeding the limit indicates that the working conditions have changed. -1 The wave number band spectrum was analyzed for consistency. Several spectra that were continuous in time were selected as windows for mean normalization and then standard deviation was calculated. The window width f = 5, the number of continuous measurement windows l = 20, and the standard deviation threshold δ = 0.04% were set. The results are shown in Figure 2. Figure 5It can be seen that the operating conditions changed from the 9th to the 16th sampling period, and then the spectrum consistency was good, and the standard deviations of the 20 consecutive windows were all within the standard deviation threshold, indicating that the operating conditions entered a new stable state at this time.

Claims

1. A water content early warning method in near infrared online analysis of distillate oil products in atmospheric and vacuum unit, characterized in that The following steps are involved: (1) Sampling multiple groups of online analysis of near-infrared spectral data and their properties; (2) Analyze the near-infrared spectrum and select a1-a2 cm that is smooth and has a significant baseline drift caused by water content -1 Wave number band spectral data X1 is used for early warning of water content in oil products. The absorption peaks of oil components b1 to b2 cm are selected. -1 Wave number band spectrum data X2 is used for property detection deviation alarm; (3) The spectral data is feature decomposed to obtain the principal component subspace and the residual subspace. The statistic SPE measures the change of the residual space, and the statistic T 2 To measure the change of principal component space, we selected samples with normal working conditions in X1 and used principal component analysis to establish the water content early warning model M1, and obtained SPE and T 2 Control limits for SPE α1 and (4) Compare the properties calculated by modeling the near-infrared spectral data of oil products with the properties obtained by chemical analysis, and use the deviation percentage β as a threshold to divide the spectra in X2 into two categories, one of which is spectral data Z1 with tolerable property deviation, and the other is spectral data Z2 with intolerable property deviation; (5) Use the spectral data Z1 to perform principal component analysis to establish the property deviation alarm model M2 and obtain the control limit SPE α2 and Add model M2 to the working condition library; (6) Use model M1 and model M2 simultaneously to monitor oil properties online; (7) According to the statistics SPE1 of the spectral data monitored by model M1, and the statistics SPE2 of the spectral data monitored by model M2, Perform different actions: If the statistic SPE1 of the monitored spectral data, and SPE2, If the oil does not exceed the limit, the oil to be tested does not contain water, and the process goes to step (6); If SPE1, If the limit is exceeded, SPE2, If the limit is not exceeded, the warning state is entered. At this time, the oil to be tested contains water, but the impact on the near-infrared oil property detection is within the tolerable range, and the process goes to step (6); If SPE1, and SPE2, If both exceed the limit, the alarm state is entered. At this time, the influence of the water component of the oil product to be tested on the near-infrared oil product property detection exceeds the tolerance range, the online monitoring device is terminated, the water and impurities accumulated in the optical pool of the online spectrometer are cleaned, the online monitoring device is restarted, and the process goes to step (6); If SPE1, Not exceeding the limit and SPE2, If an over-limit occurs, the operating condition changes. At this time, the oil does not contain water, but the oil composition has changed, resulting in a significant change in the absorption peak of the oil spectrum X2. The consistency of the spectral data X2 after the operating condition changes is calculated. When the consistency meets the requirements, it means that the operating condition tends to be stable. Determine whether the operating condition already exists in the operating condition library. If the operating condition does not appear in the library, go to step (4).

2. The method for early warning of water content in near infrared online analysis of distillate oil products in a atmospheric and vacuum unit according to claim 1, characterized in that a1~a2cm -1 The wave number range is 4705~5550cm -1 , b1~b2cm -1 The wave number range is 5639~5921cm -1 .

3. The method for early warning of water content in near infrared online analysis of distillate oil products in a atmospheric and vacuum unit according to claim 1, characterized in that The principal component analysis method is modeled in the following way: (1) For a sample matrix X with m rows and n columns, where n is the number of variables and m is the number of samples, perform column standardization on X, calculate the covariance matrix S of X, and perform eigendecomposition. The results are as follows: Where P = [p1…p n ] is the load matrix, and the eigenvalues ​​in the diagonal matrix satisfy λ1≥λ2≥…≥λ n ; (2) Based on the above formula, the expression of X is derived: T=XP=[t1……t n ] is the score matrix, and the first k variables with the largest score vector are taken as the main elements: in is the principal component matrix composed of the extracted principal components, and E is the residual matrix; (3) Establish T in the principal element space 2 Statistics, establish SPE statistics in residual space: SPE=||(I-PP T )x|| 2 T 2 =x T PΛ -1 P T x Among them, SPE α is the upper limit of the SPE statistic, It is T 2 The upper limit of statistics, λ j is the jth eigenvalue of the covariance matrix of the sample matrix X, c α is the upper 1-α quantile of the Gaussian distribution, α is the test significance level, n is the number of samples, and F(k,nk,α) is the critical value of the F distribution with k and nk degrees of freedom and a test significance level of α.

4. The method for early warning of water content in near infrared online analysis of distillate oil products in a atmospheric and vacuum unit according to claim 3, characterized in that The number of pivots k is determined as follows:

5. The method for early warning of water content in near infrared online analysis of distillate oil products in a atmospheric and vacuum unit according to claim 3, characterized in that The test significance level is α=1%.

6. The method for early warning of water content in near infrared online analysis of distillate oil products in a atmospheric and vacuum unit according to claim 1, characterized in that The threshold value of oil property deviation caused by spectral distortion is β=5%. When the spectral measurement property deviates by more than 5% from the test analysis result, the spectral detection result has no quantitative analysis value.

7. The method for early warning of water content in near infrared online analysis of distillate oil products in a atmospheric and vacuum unit according to claim 1, characterized in that Spectral consistency testing is performed as follows: (1) Setting a window, the window continuously stores f spectral data, and performing mean normalization on the spectral data X to obtain spectral data E; (2) Calculate the standard deviation of the spectral data E and take the mean to obtain the comprehensive standard deviation index s: Where n is the number of variables, e ij is the jth variable value of the ith spectral data in the window, u j is the mean of the jth variable in the window, s j is the standard deviation of the j-th variable; (3) If the standard deviation of l windows is lower than the threshold value δ, it indicates that the spectrum consistency is good and the working condition is stable.

8. The method for early warning of water content in near infrared online analysis of distillate oil products in a atmospheric and vacuum unit according to claim 7, characterized in that The window width of the spectrum consistency detection is f=5, the number of consecutive measurement windows is l=20, and the standard deviation threshold is δ=0.04%.