CN107220705A - Atmospheric and vacuum distillation unit Atmospheric Tower does Forecasting Methodology - Google Patents
Atmospheric and vacuum distillation unit Atmospheric Tower does Forecasting Methodology Download PDFInfo
- Publication number
- CN107220705A CN107220705A CN201610163932.7A CN201610163932A CN107220705A CN 107220705 A CN107220705 A CN 107220705A CN 201610163932 A CN201610163932 A CN 201610163932A CN 107220705 A CN107220705 A CN 107220705A
- Authority
- CN
- China
- Prior art keywords
- data
- variable
- atmospheric
- value
- tower
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
Abstract
Forecasting Methodology is done the present invention relates to a kind of atmospheric and vacuum distillation unit normal pressure column overhead, mainly solves to there is no the problem of Atmospheric Tower does flexible measurement method in the prior art.The present invention does Forecasting Methodology by using a kind of atmospheric and vacuum distillation unit Atmospheric Tower, for by logging in forecasting system, carrying out the prediction that Atmospheric Tower is done；The forecasting system is installed on server, server of the server by netting twine respectively with real-time dataBase system, LIMS systems is connected, client is that the technical scheme for the computer and mobile terminal having permission preferably solves above mentioned problem, available in atmospheric and vacuum distillation unit.
Description
Technical field
Forecasting Methodology is done the present invention relates to a kind of atmospheric and vacuum distillation unit Atmospheric Tower.
Background technology
In atmospheric and vacuum distillation unit, it is the normal major quality controlling index for pushing up product that the top oil of atmospheric tower, which is done, and it is main anti-
The weight of extraction oil product is reflected, so the quality of its control is not only related to the extracting rate of atmospheric tower crude oil, while influenceing below
Process.Currently, for doing, measured value can be also provided in real time without suitable instrument, most refineries are also to rely on
The manual analysis value in laboratory.For manual analysis, obtained a result from spot sampling to laboratory assay, be entered into LIMS systems again
System, the time is considerably long, about 1h-2h；In addition, the cycle of manual analysis is generally every 4 hours or 8 hours once.Cause
This, temporal delayed very serious, tower top under operating condition before can only probably having been understood by such a procedure
The quality of product, can not realize the real-time direct control to product quality at all.
In order to solve the above problems, the research of some this respects is academicly related to, but is existed in actual applications pre-
The problem of surveying not high precision, poor robustness.In process control, the method for developing many hard measurements, such as a kind of often top oil is dry
Point online soft sensor method (application number 201110198455.5), it is a kind of it is online determine Atmospheric Tower naphtha quality index it is soft
Measuring method (application number 200710171116.1), estimates to do value using a variety of mathematical modelings.But do not have also
The application patent of hard measurement is done on Atmospheric Tower.
The basic thought of hard measurement is that Theory of Automatic Control is combined with production process knowledge, appliance computer
Technology, for being difficult to measurement or temporarily immeasurable significant variable (or be active variable), selection other is easy
The variable (or be auxiliary variable) of measurement, infers and estimates by constituting certain mathematical relationship, is replaced with software firmly
Part (sensor) function.The response of this kind of method is rapid, can continuously provide active variable information, and with invest it is low, safeguard
The advantages of maintaining simple.
The content of the invention
The technical problems to be solved by the invention are the problem of still Atmospheric Tower do flexible measurement method in the prior art, are carried
Forecasting Methodology is done for a kind of new atmospheric and vacuum distillation unit Atmospheric Tower.This method is used in atmospheric and vacuum distillation unit, with test data
Accurately, measurement result more closing to reality the advantages of.
To solve the above problems, the technical solution adopted by the present invention is as follows：A kind of atmospheric and vacuum distillation unit Atmospheric Tower is done pre-
Survey method, for by logging in forecasting system, carrying out the prediction that Atmospheric Tower is done；The forecasting system is installed on server
On, server of the server by netting twine respectively with real-time dataBase system, LIMS systems is connected, and client is the electricity having permission
Brain and mobile terminal；The job step of forecasting system is as follows：
1) auxiliary variable of selection
According to execute-in-place industry control experience, it is considered to which actual done on tower top influences larger related auxiliary variable, including
The feeding temperature of atmospheric tower, often top capacity of returns, tower top pressure, tower top temperature, tower bottom steam amount and overhead reflux ratio；
2) unruly-value rejecting of auxiliary variable initial data
Using the method for moving window median filter, the abnormity point of the single process variable of ONLINE RECOGNITION, rejecting abnormalities
Value, formula is as follows：
MAD=1.4826*median | Xi-X*|}
|Xi-X*|>t*MAD
Wherein, median is the function for seeking median, X*It is the middle position of data, 1.4826 be coefficient, and threshold values t=3 is moved
The size of dynamic window takes 11 points, and the median of rejecting is filled up using the median calculated；
Initial data is after unruly-value rejecting, hence it is evident that the data for deviateing annex moment measured value are removed；
3) auxiliary variable N6ise deletion
(1) the preliminary denoising of wavelet method
Initial data is decomposed into HFS and low frequency part by the wavelet decomposition of measurement signal, the reflection of its HFS
Noise jamming, and low frequency part reflection be signal actual value；
From haar small echos, original uni-variate signal is decomposed into HFS and low frequency part using following formula：
In formula, d is scale coefficient, and β is wavelet coefficient, and G and H are high pass and low pass resolution filter respectively, and l joins for the time
Number；
Decomposition scale n=3, HFS is all filtered out, and is reconstructed with following formula：
In formula, G*And H*For high pass and low-pass reconstruction filters；
Data after reconstruct do not contain the HFS of initial data, that is, eliminate the noise of HFS so that
Data for soft instrument more accurately reflect the actual value of instrument；
(2) principle component analysis depth denoising
Using principle component analysis by the auxiliary variable data after preliminary denoising, unusual service condition identification is carried out, it is different to reject
Normal influence of the operating mode to modeling, realizes depth denoising；
Data are standardized as the following formula first：
Wherein,
Wherein,For data after standardization, xiFor initial data,For the average value of initial data, s is standard deviation；
Data after standardization are decomposed as the following formula：
In formula, pivot number k=5, this principal component model is as follows in the square error at i moment：
In formula, XijFor the measured value of j-th of input variable of i moment,For the principal component model of j-th of data variable of i moment
Predicted value, T2The control limit of statistic is calculated as follows using F distributions：
Wherein, Fk,m-1,aInsolation level a is corresponded to, the free degree is the F distribution critical values under the conditions of k, m-1；
Insolation level a=0.05, the free degree k=5, m are the width of moving window, take the data of half an hour：1/
15s, m=120, and to SPE and T2Draw the control that cumulative distribution is 95% to limit, as SPE or T2Prescribed a time limit now beyond 95% control
Operating mode will be identified that unusual service condition, its data, which is not used in, sets up soft-sensing model；
4) determination of the active variable relative to the lag time of auxiliary variable
Lag time is determined using genetic algorithm, method is specific as follows：
Genetic algorithm input variable is as follows：
N=[N1,N2,…,Nj] j=1,2 ..., m
Wherein, NjFor the lag time of j-th of input variable, m is auxiliary variable number；
Genetic algorithm object function is as follows：
Wherein, yiIt is the offline laboratory values of leading variable,It is the folding cross validation predicted value of GRNN models 5, n is training sample
Number；
In soft-sensing model is set up, m=6 is taken, lag time scope is Nj=0~60min, due to NjValue be just
Integer, is converted into length and is 6 binary system to calculate；The Population Size of genetic algorithm is 200, random initializtion population, iteration
Number of times is 500, and crossover probability is 0.4, and mutation probability is 0.2；
5) flexible measurement method；
Using generalized regression nerve networks (Generalized regression neural network, GRNN) to normal
The soft sensor modeling that pressure tower Atmospheric Tower is done, GRNN network structures are constituted by four layers, respectively input layer, mode layer, summation
Layer and output layer；Wherein, input layer number is 6, and mode layer neuron number is the number of training sample, the nerve of output layer
First number is equal to 1.Mode layer neural transferring function is：
Summation layer neuron transmission function be：
The transmission function of output layer neuron is：
The expansion rate Spread of RBF during GRNN model trainings is defined as by 5 folding cross validation methods
Spread=0.2；
6) system algorithm technology path
7) data-interface is developed
In order to obtain the creation data of actual device, a variety of data acquisition interfaces are developed, from a variety of Petrochemical Enterprises main flows
Real-time data base gathers the data of auxiliary variable, the need for meeting various field conduct environment；Meanwhile, develop ODBC interfaces
Connect LIMS systems (the Laboratory Information Management System, laboratory calibration system of enterprise
System), the online data for obtaining active variable realize that Atmospheric Tower does the inspection and amendment of prediction forecasting system algorithm.
The method of this patent adds unruly-value rejecting, the N6ise deletion to auxiliary variable initial data, reduces and even avoids
Unnecessary interference, makes sample data more accurate；Lag time of the active variable relative to auxiliary variable is considered, meets work
The reality of industry operation, active variable predicts the outcome closer to reality；The on-line monitoring system of exploitation, with various field data
Acquisition interface, strong adaptability achieves preferable technique effect.
Brief description of the drawings
Fig. 1 is that atmospheric and vacuum distillation unit Atmospheric Tower does forecasting system algorithm logic figure.
Fig. 2 is hardware distribution figure.
In Fig. 2,1 is real-time data base；2 be auxiliary variable；3 be LIMS databases；4 be active variable；5 be fire wall；
6 be that atmospheric and vacuum distillation unit Atmospheric Tower does forecasting system；7 be wireless router；8 be tablet personal computer；9 be office computer.
Below by embodiment, the invention will be further elaborated, but is not limited only to the present embodiment.
Embodiment
【Embodiment 1】
Using the method for the present invention, for by logging in forecasting system, carrying out the prediction that Atmospheric Tower is done；The prediction
System is installed on server, and server of the server by netting twine respectively with real-time dataBase system, LIMS systems is connected, visitor
Family end is the computer and mobile terminal having permission；The job step of forecasting system is as follows：
1) auxiliary variable of selection
According to execute-in-place industry control experience, it is considered to which actual done on tower top influences larger related auxiliary variable, including
The feeding temperature of atmospheric tower, often top capacity of returns, tower top pressure, tower top temperature, tower bottom steam amount and overhead reflux ratio.
2) unruly-value rejecting of auxiliary variable initial data
Outlier refers to that the value at a certain moment in process variable measurement deviates considerably from the measured value of other adjacent moments.Outlier
It is due to that caused by measuring apparatus error or noise, its value can not reflect real operating mode.It will be reduced to soft if do not rejected
The precision of instrument model.This patent use moving window median filter method, the single process variable of ONLINE RECOGNITION it is different
Chang Dian, rejecting abnormalities value.Formula is as follows：
MAD=1.4826*median | Xi-X*|}
|Xi-X*|>t*MAD
Wherein, median is the function for seeking median, X*It is the middle position of data, 1.4826 be coefficient, threshold values t=3.Move
The size of dynamic window takes 11 points, and the median of rejecting is filled up using the median calculated.
Initial data is after unruly-value rejecting, hence it is evident that deviateing the data of annex moment measured value can be removed.
3) auxiliary variable N6ise deletion (small echo, PCA)
Noise is the random error of generally existing in measurement data, its value Normal Distribution.Noise has to measurement data
Obvious influence, makes measured value deviate actual value.
(2) the preliminary denoising of wavelet method
Initial data can be decomposed into HFS and low frequency part by the wavelet decomposition of measurement signal, and its HFS is anti-
Reflect the noise jamming for being, and low frequency part reflection be signal actual value.
From haar small echos, original uni-variate signal is decomposed into HFS and low frequency part using following formula.
In formula, d is scale coefficient, and β is wavelet coefficient, and G and H are high pass and low pass resolution filter respectively, and l joins for the time
Number.
Decomposition scale n=3, HFS is all filtered out, and is reconstructed with following formula.
In formula, G*And H*For high pass and low-pass reconstruction filters.
Data after reconstruct do not contain the HFS of initial data, that is, eliminate the noise of HFS so that
Data for soft instrument more accurately reflect the actual value of instrument.
(2) principle component analysis depth denoising
The related unusual service condition of multiple auxiliary variables falls within a kind of noise, will tentatively be gone using principle component analysis (PCA)
Auxiliary variable data after making an uproar, carry out unusual service condition identification, and the influence so as to rejecting abnormalities operating mode to modeling realizes that depth is gone
Make an uproar.Data are standardized as the following formula first,
Wherein,
Wherein,For data after standardization, xiFor initial data,For the average value of initial data, s is standard deviation.
Data after standardization are decomposed as the following formula
In formula, pivot number k=5, this principal component model is as follows in the square error (SPE) at i moment
In formula, XijFor the measured value of j-th of input variable of i moment,For the principal component model of j-th of data variable of i moment
Predicted value.
T2The control limit of statistic can be calculated as follows using F distributions
Wherein, Fk,m-1,aInsolation level a is corresponded to, the free degree is the F distribution critical values under the conditions of k, m-1.
Here, insolation level a=0.05, the free degree k=5, m are the width of moving window, take the data (1 of half an hour
Individual/15s), m=120.And to SPE and T2Draw the control that cumulative distribution is 95% to limit, as SPE or T2Prescribed a time limit beyond 95% control
Operating mode now will be identified that unusual service condition, and its data, which is not used in, sets up soft-sensing model.
4) determination of the active variable relative to the lag time of auxiliary variable
Because the flow of atmospheric and vacuum distillation unit is longer, retardance of the operation with the time, auxiliary variable is in moment t1Operation,
Will be to moment t2(t2>t1) can just be reflected on leading variable, it is thus necessary to determine that leading variable is stagnant relative to auxiliary variable
Time afterwards.
This patent determines lag time using genetic algorithm, and method is specific as follows：
Genetic algorithm input variable is as follows
N=[N1,N2,…,Nj] j=1,2 ..., m
Wherein, NjFor the lag time of j-th of input variable, m is auxiliary variable number.
Genetic algorithm object function is as follows
Wherein, yiIt is the offline laboratory values of leading variable,It is the folding cross validation predicted value of GRNN models 5, n is training sample
Number.
In soft-sensing model is set up, m=6 is taken, lag time scope is Nj=0~60min.Due to NjValue be just
Integer, is converted into length and is 6 binary system (lag time that can represent 0~63min) to calculate.The population of genetic algorithm is big
Small is 200, and random initializtion population, iterations is 500, and crossover probability is 0.4, and mutation probability is 0.2.
5) flexible measurement method (GRNN)
This patent uses the soft sensor modeling that generalized regression nerve networks (GRNN) are done to atmospheric tower Atmospheric Tower.GRNN
Network structure is constituted by four layers, respectively input layer, mode layer, summation layer and output layer.Wherein, input layer number is 6, mould
Formula layer neuron number is the number of training sample, and the neuron number of output layer is equal to 1.Mode layer neural transferring function is
Summation layer neuron transmission function be
The transmission function of output layer neuron is
The expansion rate Spread of RBF during GRNN model trainings is defined as by 5 folding cross validation methods
Spread=0.2.
6) system algorithm technology path, as shown in Figure 1.
7) data-interface is developed
In order to obtain the creation data of actual device, develop a variety of data acquisition interfaces, such as API, ODBC,
WebService, OPC etc., can be from InfoPlus.21, Plant Information System, Process History
The Petrochemical Enterprises main flow such as Database real-time data base gathers the data of auxiliary variable, can meet the need of various field conduct environment
Will.
Meanwhile, the LIMS systems that ODBC interfaces connect enterprise are developed, the online data for obtaining active variable realize normal pressure
Tower top does the inspection and amendment of forecasting system algorithm.
8) hardware environment
Hardware configuration is as shown in Figure 2.Central Control Room configure a server, by netting twine respectively with real-time data base system
System, the server of LIMS systems are connected, and install and run " atmospheric and vacuum distillation unit Atmospheric Tower does forecasting system " server version.
Client can be the computer arbitrarily having permission and mobile terminal for being in enterprise's Office Network, such as smart mobile phone, flat
Plate computer etc..
9) server-side system application
9.1 activation system
Hardware is connected, starts subsystems, server is opened.
9.2 control of authority
According to the user of input, detect or select different identity to enter system.
9.3 configuration modeling：
The part mainly completes the modeling of reasoning algorithm.Auxiliary variable and the history value of active variable are collected, as sample,
Carry out Algorithm for Training.
9.4 monitorings in real time
The real-time data base of enterprise is connected, system starts real-time monitoring.
1) the auxiliary variable real time data from production scene, the instantaneous value that real-time reasoning and calculation is currently done are gathered.
2) time of data is gone out according to LIMS systems, every 4-8 hours, the change of the value that hard measurement is obtained and LIMS systems
Test value to be contrasted, in real time amendment.
10) FTP client FTP application
The client developed, existing B/S frameworks, it is easy to user in enterprise using any computer on Office Network, fits
For device management personnel；There are C/S frameworks again, enable users to check calculating details in detail, it is adaptable to flat in face of operative employee
Plate computer, the PC computers of technique person office.
The method of this patent adds unruly-value rejecting, the N6ise deletion to auxiliary variable initial data, reduces and even avoids
Unnecessary interference, makes sample data more accurate；Lag time of the active variable relative to auxiliary variable is considered, meets work
The reality of industry operation, active variable predicts the outcome closer to reality；The on-line monitoring system of exploitation, with various field data
Acquisition interface, strong adaptability achieves preferable technique effect.
Claims (1)
1. a kind of atmospheric and vacuum distillation unit Atmospheric Tower does Forecasting Methodology, for by logging in forecasting system, carrying out Atmospheric Tower and doing
The prediction of point；The forecasting system is installed on server, server by netting twine respectively with real-time dataBase system, LIMS systems
The server of system is connected, and client is the computer and mobile terminal having permission；The job step of forecasting system is as follows：
1) auxiliary variable of selection
According to execute-in-place industry control experience, it is considered to which actual done on tower top influences larger related auxiliary variable, including normal pressure
The feeding temperature of tower, often top capacity of returns, tower top pressure, tower top temperature, tower bottom steam amount and overhead reflux ratio；
2) unruly-value rejecting of auxiliary variable initial data
Using the method for moving window median filter, the abnormity point of the single process variable of ONLINE RECOGNITION, rejecting abnormalities value is public
Formula is as follows：
MAD=1.4826*median | Xi-X*|}
|Xi-X*|>t*MAD
Wherein, median is the function for seeking median, X*It is the middle position of data, 1.4826 be coefficient, threshold values t=3, Moving Window
The size of mouth takes 11 points, and the median of rejecting is filled up using the median calculated；
Initial data is after unruly-value rejecting, hence it is evident that the data for deviateing annex moment measured value are removed；
3) auxiliary variable N6ise deletion
(1) the preliminary denoising of wavelet method
Initial data is decomposed into HFS and low frequency part by the wavelet decomposition of measurement signal, and the reflection of its HFS is to make an uproar
Sound disturb, and low frequency part reflection be signal actual value；
From haar small echos, original uni-variate signal is decomposed into HFS and low frequency part using following formula：
In formula, d is scale coefficient, and β is wavelet coefficient, and G and H are high pass and low pass resolution filter respectively, and l is time parameter；
Decomposition scale n=3, HFS is all filtered out, and is reconstructed with following formula：
In formula, G*And H*For high pass and low-pass reconstruction filters；
Data after reconstruct do not contain the HFS of initial data, that is, eliminate the noise of HFS so that be used for
The data of soft instrument more accurately reflect the actual value of instrument；
(2) principle component analysis depth denoising
Using principle component analysis by the auxiliary variable data after preliminary denoising, unusual service condition identification is carried out, so as to rejecting abnormalities work
Influence of the condition to modeling, realizes depth denoising；
Data are standardized as the following formula first：
Wherein,
Wherein,For data after standardization, xiFor initial data,For the average value of initial data, s is standard deviation；
Data after standardization are decomposed as the following formula：
In formula, pivot number k=5, this principal component model is as follows in the square error at i moment：
In formula, XijFor the measured value of j-th of input variable of i moment,Predicted for the principal component model of j-th of data variable of i moment
Value, T2The control limit of statistic is calculated as follows using F distributions：
Wherein, Fk,m-1,aInsolation level a is corresponded to, the free degree is the F distribution critical values under the conditions of k, m-1；
Insolation level a=0.05, the free degree k=5, m are the width of moving window, take the data of half an hour：1/15s, m=
120, and to SPE and T2Draw the control that cumulative distribution is 95% to limit, as SPE or T2Beyond the operating mode of 95% control in limited time now
Unusual service condition is will be identified that, its data, which is not used in, sets up soft-sensing model；
4) determination of the active variable relative to the lag time of auxiliary variable
Lag time is determined using genetic algorithm, method is specific as follows：
Genetic algorithm input variable is as follows：
N=[N1,N2,…,Nj] j=1,2 ..., m
Wherein, NjFor the lag time of j-th of input variable, m is auxiliary variable number；
Genetic algorithm object function is as follows：
Wherein, yiIt is the offline laboratory values of leading variable,It is the folding cross validation predicted value of GRNN models 5, n is training sample number；
In soft-sensing model is set up, m=6 is taken, lag time scope is Nj=0-60min, due to NjValue be positive integer,
Length is converted into be 6 binary system to calculate；The Population Size of genetic algorithm is 200, random initializtion population, and iterations is
500, crossover probability is 0.4, and mutation probability is 0.2；
5) flexible measurement method；
Use generalized regression nerve networks；The soft sensor modeling done to atmospheric tower Atmospheric Tower, GRNN network structures are by four layers
Constitute, respectively input layer, mode layer, summation layer and output layer；Wherein, input layer number is 6, mode layer neuron number
For the number of training sample, the neuron number of output layer is equal to 1.Mode layer neural transferring function is：
Summation layer neuron transmission function be：
The transmission function of output layer neuron is：
The expansion rate Spread of RBF during GRNN model trainings is defined as by 5 folding cross validation methods
Spread=0.2；
6) system algorithm technology path
7) data-interface is developed
In order to obtain the creation data of actual device, a variety of data acquisition interfaces are developed, it is real-time from a variety of Petrochemical Enterprises main flows
Database gathers the data of auxiliary variable, the need for meeting various field conduct environment；Meanwhile, develop the connection of ODBC interfaces
The LIMS systems of enterprise, the online data for obtaining active variable, realize that Atmospheric Tower does the inspection of forecasting system algorithm with repairing
Just.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610163932.7A CN107220705B (en) | 2016-03-22 | 2016-03-22 | Atmospheric tower top dry point prediction method for atmospheric and vacuum device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610163932.7A CN107220705B (en) | 2016-03-22 | 2016-03-22 | Atmospheric tower top dry point prediction method for atmospheric and vacuum device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107220705A true CN107220705A (en) | 2017-09-29 |
| CN107220705B CN107220705B (en) | 2020-10-09 |
Family
ID=59927211
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610163932.7A Active CN107220705B (en) | 2016-03-22 | 2016-03-22 | Atmospheric tower top dry point prediction method for atmospheric and vacuum device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107220705B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108152791A (en) * | 2017-12-06 | 2018-06-12 | 南京航空航天大学 | Radio cheating signal framing system based on GRNN neural networks |
| CN109658988A (en) * | 2018-12-20 | 2019-04-19 | 中海石油炼化有限责任公司 | A kind of Hydrobon catalyst performance prediction method |
| CN111592907A (en) * | 2020-05-18 | 2020-08-28 | 南京富岛信息工程有限公司 | Online analysis and detection method for oil dry point of atmospheric tower top |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2281269A1 (en) * | 1998-09-01 | 2000-03-01 | A.C. Analytical Controls Holding B.V. | A method and apparatus for predicting a distillation temperature range of a hydrocarbon-containing compound |
| CN101169387A (en) * | 2007-11-28 | 2008-04-30 | 华东理工大学 | Soft sensing method for on-line determination of atmospheric tower top naphtha quality index |
| CN101201331A (en) * | 2007-11-28 | 2008-06-18 | 华东理工大学 | Soft measuring method for on-line determining petroleum naphtha quality index on top of primary tower |
| EP2416153A1 (en) * | 2010-08-02 | 2012-02-08 | ENI S.p.A. | Method for predicting the quality and yields of a crude oil |
| CN102880905A (en) * | 2011-07-15 | 2013-01-16 | 中国石油天然气股份有限公司 | Atmospheric overhead oil dry point on-line soft measurement method |
| CN104965967A (en) * | 2015-03-26 | 2015-10-07 | 华东理工大学 | Yield real-time prediction method for atmospheric and vacuum distillation unit |
-
2016
- 2016-03-22 CN CN201610163932.7A patent/CN107220705B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2281269A1 (en) * | 1998-09-01 | 2000-03-01 | A.C. Analytical Controls Holding B.V. | A method and apparatus for predicting a distillation temperature range of a hydrocarbon-containing compound |
| CN101169387A (en) * | 2007-11-28 | 2008-04-30 | 华东理工大学 | Soft sensing method for on-line determination of atmospheric tower top naphtha quality index |
| CN101201331A (en) * | 2007-11-28 | 2008-06-18 | 华东理工大学 | Soft measuring method for on-line determining petroleum naphtha quality index on top of primary tower |
| EP2416153A1 (en) * | 2010-08-02 | 2012-02-08 | ENI S.p.A. | Method for predicting the quality and yields of a crude oil |
| CN102880905A (en) * | 2011-07-15 | 2013-01-16 | 中国石油天然气股份有限公司 | Atmospheric overhead oil dry point on-line soft measurement method |
| CN104965967A (en) * | 2015-03-26 | 2015-10-07 | 华东理工大学 | Yield real-time prediction method for atmospheric and vacuum distillation unit |
Non-Patent Citations (3)
| Title |
|---|
| YAFEN LI 等: "Soft Sensing Based on LS-SVM and Its Application to a Distillation Column", 《SIXTH INTERNATIONAL CONFERENCE ON INTELLIGENT SYSTEMS DESIGN AND APPLICATIONS》 * |
| 张笑天 等: "基于多神经网络模型的石脑油干点软测量", 《控制工程》 * |
| 金思毅 等: "常减压装置常压塔塔顶汽油干点的软测量", 《化工进展》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108152791A (en) * | 2017-12-06 | 2018-06-12 | 南京航空航天大学 | Radio cheating signal framing system based on GRNN neural networks |
| CN109658988A (en) * | 2018-12-20 | 2019-04-19 | 中海石油炼化有限责任公司 | A kind of Hydrobon catalyst performance prediction method |
| CN111592907A (en) * | 2020-05-18 | 2020-08-28 | 南京富岛信息工程有限公司 | Online analysis and detection method for oil dry point of atmospheric tower top |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107220705B (en) | 2020-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112101480B (en) | Multivariate clustering and fused time sequence combined prediction method | |
| Mogheir et al. | Characterizing the spatial variability of groundwater quality using the entropy theory: I. Synthetic data | |
| CA2640725C (en) | Methods, systems, and computer-readable media for fast updating of oil and gas field production models with physical and proxy simulators | |
| CN104182642B (en) | A kind of fault detection method based on rarefaction representation | |
| CN103996077A (en) | Electric equipment fault forecasting method based on multi-dimension time sequence | |
| CN109992921B (en) | On-line soft measurement method and system for thermal efficiency of boiler of coal-fired power plant | |
| CN109376401A (en) | A kind of adaptive multi-source information preferably with the mechanical method for predicting residual useful life that merges | |
| CN109471698B (en) | System and method for detecting abnormal behavior of virtual machine in cloud environment | |
| CN111275255A (en) | Construction method of concrete dam deformation monitoring and forecasting model | |
| JP2023518862A (en) | Systems and methods for developing and deploying well models that predict wax/hydrate accumulation for well optimization | |
| CN113036913B (en) | Method and device for monitoring state of comprehensive energy equipment | |
| CN107220705A (en) | Atmospheric and vacuum distillation unit Atmospheric Tower does Forecasting Methodology | |
| CN109976311A (en) | Single order fixed set point control system sensor fault diagnosis method and diagnostic system | |
| CN106874676B (en) | State evaluation method for electric energy metering device | |
| CN115438726A (en) | Device life and fault type prediction method and system based on digital twin technology | |
| Kurt | Determination of the most appropriate statistical method for estimating the production values of medium density fiberboard | |
| CN112016241B (en) | Soft measurement modeling method based on mLASSO-MLP model | |
| CN107220393A (en) | The normal line of atmospheric and vacuum distillation unit atmospheric tower does Forecasting Methodology | |
| Requeijo et al. | Implementation of the statistical process control with autocorrelated data in an automotive manufacturer | |
| CN107220252A (en) | The normal line prediction of Flash method of atmospheric and vacuum distillation unit atmospheric tower | |
| CN104200269A (en) | Real-time fault diagnosis method based on online learning minimum embedding dimension network | |
| CN109635358A (en) | A kind of unit fault detection method based on sliding window Multiscale Principal Component Analysis | |
| CN113468720B (en) | Service life prediction method for digital-analog linked random degradation equipment | |
| Adedotun et al. | Modelling and forecasting climate time series with state-space model | |
| CN107220392A (en) | The normal point prediction method of a line 10% of atmospheric and vacuum distillation unit atmospheric tower |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |