CN101727609B - Pyrolyzate yield forecasting method based on support vector machine - Google Patents
Pyrolyzate yield forecasting method based on support vector machine Download PDFInfo
- Publication number
- CN101727609B CN101727609B CN200810225363XA CN200810225363A CN101727609B CN 101727609 B CN101727609 B CN 101727609B CN 200810225363X A CN200810225363X A CN 200810225363XA CN 200810225363 A CN200810225363 A CN 200810225363A CN 101727609 B CN101727609 B CN 101727609B
- Authority
- CN
- China
- Prior art keywords
- yield
- vector machine
- machine model
- svms
- supporting vector
- 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.)
- Active
Links
Abstract
The invention discloses a pyrolyzate yield forecasting method based on a support vector machine (SVM). Aiming at the problems that ethylene plants need to forecast the pyrolyzate yield of different pyrolysis raw materials under different operation conditions and the operation conditions need to be optimized, the invention takes the physical property of known pyrolysis raw materials, pyrolysis technological conditions and product yield as a basis of sample data to build an SVM model estimated by a regression function as well as utilize the SVM model to forecast the pyrolysis yield. The method not only can forecast the product yield under different pyrolysis raw materials and pyrolysis technological conditions, but also can optimize the pyrolysis technological conditions so as to obtain maximum economic benefit.
Description
Technical field
The present invention relates to the technology of ethylene industry cracking, more particularly, the present invention relates to predict the method for Pyrolyzate yield.
Background technology
Ethene is the most important basic material of chemical industry, is that basis or derivative series chemical products are widely used in the national economy every field with ethene, relates to huge economic interests.Along with domestic fast development of national economy, also increasing to the demand of ethene, domestic ethene total production reached 1024.8 ten thousand tons in 2007, but can't satisfy domestic demand to ethene and ethylene derivative, needed a large amount of ethene of import and derived product thereof.In order to satisfy domestic great demand to ethene, this year is domestic has carried out second and third and has taken turns the expansion of ethylene engineering, and by 2010, China's ethene total productive capacity will reach 1,702 ten thousand tons/year.At present, so that thermal cracking/the steam cracking mode is produced, and China's ethene basically all is to produce in this way to the ethene in the whole world 99% through tube cracking furnace.
The raw materials for production major part of tube cracking furnace is all come the lighter hydrocarbons of oil gas field and the portioned product of oil refining apparatus, like naphtha, diesel oil, hydrogenation tail oil.Because domestic hydrocarbon resources is relatively poorer, the main dependence on import of oil, therefore domestic cracking stock is main with the product of oil refining apparatus mainly, like naphtha.In recent years crude oil price is higher always, causes the crude oil wide material sources, and crude oil property changes greatly, causes the rerum natura of cracking stock to change also very frequent.From the present condition of production of domestic olefin plant; Although having carried out enlarging, transforms most of olefin plant tube cracking furnace; And oil refining apparatus productive capacity does not obtain corresponding increase, causes the cracking stock degradation, and is under-supply; Compelled other oil products that replenish are made cracking stock, like hydrocracking diesel oil etc.Because domestic most pyrolysis furnaces all do not have the foundation of thermal cracking model as the pyrolysis furnace operation with reference, thus this operation has brought sizable difficulty to pyrolysis furnace.
The yield of thermal cracking products such as prediction tube cracking furnace ethene has three kinds of methods: the dynamic (dynamical) empirical model of heat scission reaction, the model of half, mechanism model.External Pyrolyzate yield research is started in eighties of last century thirties, because this technology relates to the core technology of tube cracking furnace, often this technical research process is strictly maintained secrecy, like LUMMUS, TECHNIP, S&W, LINDE.Though external pyrolysis furnace patent merchant also carries business prototype software, price is very expensive, limits morely, and effect is still to be tested.And the domestic also untapped similar business software that goes out.
SVMs (SVM) is a kind of intellectual technology; Its basic thought is to shine upon through the linear/non-linear of prior selection input vector X is mapped to a high-dimensional feature space Z, then structure optimized data qualification/data regression hyperplane in this feature space.SVM is a kind of nonparametric machine learning method, has realized the design philosophy of structural risk minimization in the machine learning, has solved to perplex many problems of machine learning method in the past.Like the problem of model selection of neural network, cross study and owe study, non-linear and problems such as dimension disaster problem and local minimum point, all in SVM, obtained solving preferably.SVM is applicable to and solves problem insoluble with the traditional mathematics model method, uncertainty that its special feature is the processing power to small sample, can solve training sample practical problems on the low side in the cracking just.Modeling technique with traditional is compared; The SVM model only needs the parameter and the rational training data of related system; Can avoid the many difficult problems in the heat scission reaction kinetic model performance history like this, for prediction pyrolysis product yield provides a new approach with optimization.
Through the SVM model; Change to different cracking stocks in the ethylene plant; Can dope the pyrolysis product yield of pyrolysis furnace under the different operating condition; Can seek out production target product (like ethene+propylene, ethene+propylene+butadiene) than operation conditions optimization, improve the target product yield of pyrolysis furnace, thereby improve the economic benefit of manufacturing enterprise.
Summary of the invention
The present invention is directed to the ethylene plant and need predict pyrolysis product yield and the problem of operation conditions optimization of different cracking stocks under the different operating condition; Propose a kind of Forecasting Methodology of industrial pyrolysis furnace Pyrolyzate yield, it is based on SVMs (SVM) data regression models.
Concrete technical scheme is following:
The Forecasting Methodology of the Pyrolyzate yield based on SVMs of the present invention is to be sample with thermal cracking raw material rerum natura, thermal cracking process condition and product yield data; The supporting vector machine model that foundation utilizes regression function to estimate utilizes supporting vector machine model to predict the Pyrolyzate yield of cracking stock again.
The Forecasting Methodology of said Pyrolyzate yield based on SVMs may further comprise the steps:
(1) sets up sample: collect the different cracking stock of rerum natura; To its Physical Property Analysis record; Then cracking stock and thinning agent are fed in the pyrolysis experiment device; Under different technology conditions, carry out heat scission reaction, analyze and calculate Pyrolyzate yield, the data of sample are made up of cracking stock rerum natura, thermal cracking process condition and pyrolysis product yield;
(2) set up supporting vector machine model: according to cracking stock rerum natura, thermal cracking process condition and pyrolysis product yield, coding, said program is made up of database, SVMs training program and predictor; Database adopts the database of open-source standard interface, with the sample data input database, for the SVMs training program provides the learning training sample; The support vector training program adopts the standard C language exploitation, and this program principle is based on the principle of SVMs; The SVMs training program is called open-source standard interface database; Sample data is imported in the array in the SVMs training program; The support vector training program is carried out learning training to sample data, sets up pattern-recognition with the rule to sample, forms supporting vector machine model; Wherein cracking stock rerum natura and thermal cracking process condition are as the input variable of said model, and Pyrolyzate yield is as output variable;
(3) pyrolysis product yield prediction: predictor calls supporting vector machine model, and cracking stock rerum natura and thermal cracking process condition data are input in the input variable of supporting vector machine model, dopes Pyrolyzate yield through calculating.
More particularly, the Forecasting Methodology of said Pyrolyzate yield based on SVMs may further comprise the steps:
(1) set up sample data: collect known cracking stock rerum natura, thermal cracking process condition and pyrolysis product yield as sample data, select sample data over half as the training sample data at random, the samples remaining data are as the forecast sample data;
(2) input database: with forming sample database in the training sample data input database, cracking stock rerum natura and thermal cracking process condition are as input variable, and the pyrolysis product yield is as output variable;
(3) set up supporting vector machine model: set up SVMs training program based on kernel function; Input variable and output variable form mapping relations through supporting vector machine model; Utilize the SVMs training program that the training sample data are carried out learning training, obtain N support vector X through study and training
i *, i=0,1 ..., N, thus supporting vector machine model formed:
Wherein,
The expression support vector, Y
iThe product yield of expression support vector, α
iThe coefficient of representing i support vector, X are the rerum natura and the thermal cracking process conditions of input, and Y (X) representes yield, and K () is the kernel function of SVMs, are selected from a kind of in Gaussian function, polynomial function, lienar for function and the RBF.
(4) set up the supporting vector machine model information database: with parameter information, support vector information and each coefficient storage of the supporting vector machine model of setting up in said database; Parameter information comprises radially basic kernel function type, gamma parameter and support vector number, and support vector information comprises the numerical value of each dimension of support vector;
(5) prediction pyrolysis product yield: in the supporting vector machine model that the raw material rerum natura in the forecast sample data and thermal cracking process condition entry are set up, calculate predicting the outcome of pyrolysis product yield through supporting vector machine model;
(6) confirm supporting vector machine model: what comparison step (5) obtained predicts the outcome and experimental result; When surpassing acceptable scope with the deviation of experimental result if predict the outcome; For example deviation surpasses 10%, changes the SVMs kernel function or SVMs kernel function parameter values is regulated, and then train again and predict; Until the deviation of predicted value and experimental result within the acceptable range; For example, determine supporting vector machine model like this, utilize determined this supporting vector machine model prediction cracking stock Pyrolyzate yield less than 10%.
Preferably; Said thermal cracking raw material rerum natura can be density, grace formula distillation data, cracking stock group composition, hydrogen richness or carbon content or ratio of carbon-hydrogen, carboloy residue, molecular weight, correlation index BMCI and refraction index, and perhaps said thermal cracking raw material rerum natura can be the composition of cracking stock.Cracking stock is divided into liquid charging stock and lighter hydrocarbons, and the liquid cracking stock is main with naphtha, hydrogenation tail oil, diesel oil mainly, and lighter hydrocarbons are mainly ethane, propane, butane and their potpourri.For the popularity that guarantees that model is used, select the big cracking stock of rerum natura scope to make an experiment as far as possible.For the liquid cracking stock, the main rerum natura that analyze has density, grace formula distillation data, cracking stock group composition, for lighter hydrocarbons, can analyze its detailed composition.For the selection of these rerum naturas, to guarantee to have nothing to do between each rerum natura as far as possible.
Preferably, said thermal cracking process condition comprises throwing oil mass, boiler tube temperature in, coil outlet temperature, boiler tube inlet pressure, boiler tube top hole pressure, dilution ratio and the residence time.In order to guarantee the popularity of application of model, the heat scission reaction process conditions are wide as far as possible.
Preferably, said product yield is the yield of hydrogen, methane, ethane, ethene, acetylene, propane, propylene, propine, butane, butylene, butadiene, benzene,toluene,xylene, ethylbenzene, pyrolysis gasoline, pyrolysis gas oil and the fuel oil of cracking generation.
Said thinning agent is generally water, also can select nitrogen or inert gas.The experimental provision of pyrolysis experiment can be lab scale, pilot scale or commercial plant.Cracking stock after heat scission reaction, yield that can its thermal cracking products of analytical calculation.The data of cracking stock rerum natura, thermal cracking process condition and corresponding pyrolysis product yield have just constituted the data of sample.
The database of sample can adopt EXCEL, ACCESS, SOL SEVER etc., and this database provides the data of wanting learning training for the SVMs training program; The support vector training program adopts the standard C language exploitation, and the principle of SVMs training program is based on the principle of the inventor Vapnik proposition of SVMs.The function of SVMs training program is through analysis, calculating and regular pattern identification to the training sample data, and has formed the supporting vector machine model of similar regression formula.Predictor then is to import corresponding parameter toward vector machine model, and like the rerum natura and the thermal cracking process condition of cracking stock, vector machine model can calculate the cracking stock Pyrolyzate yield according to the condition of input.For kernel function and parameters of choice, be as the criterion with predicted value and experiment value error, select less kernel function of those errors and parameter thereof.
Each sample all comprises the yield of character, pyrolysis furnace operating procedure condition and the corresponding thermal cracking products of cracking stock.The character of cracking stock and pyrolysis furnace operating conditions are input layer, and the yield of thermal cracking products is an output layer.The quantity of sample preferably is at least input layer and output layer number sum ten times.The selected rerum natura of different cracking stocks is different; The liquid cracking stock is formed complicated; Therefore its rerum natura is generally density, grace formula distillation data, cracking stock group composition, for light hydrocarbon feedstocks, because it consists of ethane, propane, butane; Component simply can be analyzed, and therefore selects its detailed component.Preferably have nothing to do between the cracking stock rerum natura.Pyrolysis furnace operating procedure condition is for throwing oil mass, the residence time, boiler tube temperature in, coil outlet temperature, boiler tube top hole pressure, WOR.Output layer is the yield of thermal cracking products, wherein should select economic worths such as hydrogen that the chemical plant relatively is concerned about, methane, ethene, propylene, butadiene, aromatic hydrocarbons thermal cracking product preferably.
According to the characteristic of cracking stock, select suitable sample to form.For lighter hydrocarbons, sample can be for its cracking stock is formed in detail, thermal cracking process condition, Pyrolyzate yield; For the liquid cracking stock; Its sample can also can be selected combinations such as density, hydrogen richness, group composition, correlation index, volume average boiling point, thermal cracking process condition, Pyrolyzate yield for density, grace formula distillation data, group composition, thermal cracking process condition, Pyrolyzate yield.
Choose after the sample, sample data is divided into two parts, most of data are as training data, and remaining few part is as predicted data.In order to guarantee the popularized type of vector machine model, 2/3rds the data of selecting sample usually immediately are as training data, and one data of its excess-three branchs are used as predicted data.Data characteristic according to training; Select the kernel function and the parameter of suitable SVMs, let SVMs fully the inherent law of training data learnt and discern, after the kernel function of having selected SVMs and parameter; Predicted data brought in the supporting vector machine model predict; If most predicted values and experiment value deviation are lower than 10%, this just explains that supporting vector machine model is qualified, can be applied to actual production.If deviation is bigger, the kernel function and the parameter of supporting vector machine model are adjusted, perhaps adjust sample, until error in 10%.
In addition; In supporting vector machine model, owing to exist mapping relations, the process conditions of the thermal cracking when we can find out certain thermal cracking products maximal value through this mapping relations between thermal cracking process condition and the Pyrolyzate yield; Pyrolysis product such as ethene, propylene, butadiene, aromatic hydrocarbons etc.; Process conditions are generally coil outlet temperature, perhaps utilize the foreseeable function of supporting vector machine model, through utilizing mathematical method; Be worth most search like numerical value, process conditions and Pyrolyzate yield when can seek out cracking stock target product maximum.For the ethylene plant, this thermal cracking target product is generally the yield maximal value of ethene+propylene, perhaps ethene+propylene+butadiene yield maximal value.Therefore, change frequent ethylene production factory for cracking stock, the SVM prediction model can be sought the best thermal cracking process condition of pyrolysis furnace, thereby improve the economic benefit of manufacturing enterprise rapidly according to the rerum natura of cracking stock.
Embodiment
Further describe the present invention below in conjunction with embodiment.Scope of the present invention is not limited by these embodiment, and scope of the present invention proposes in claims.
Embodiment 1
At first will set up sample: cracking stock is selected naphtha, and thinning agent is a water.According to the industry practice condition of production; The input layer of each sample has 12 vectors; Be the rerum natura and the thermal cracking process condition of naphtha, wherein related with thermal cracking naphtha rerum natura vector has 8: hydrogen richness, correlation index, proportion, mean average boiling point, positive structure chain hydrocarbon content, isomery chain hydrocarbon content, arene content, naphthene content; The thermal cracking process condition has 4 vectors: WOR, the residence time, coil outlet temperature (COT), boiler tube top hole pressure (COP); Output layer is set with 5 vectors, is the product yield of naphtha thermal cracking: the yield of hydrogen, methane, ethene, propylene, butadiene.
Naphtha pyrolysis experiment device adopts the simulation stove of Sinopec Beijing Research Institute of Chemical Industry cracking of ethylene research department; 16 kinds of imports and homemade naphtha have been collected altogether; Wherein density is 0.74 gram/cubic centimetre to the maximum, and minimum is 0.66 gram/cubic centimetre.According to ethylene plant's practical condition, carried out 300 test of many times altogether, wherein WOR is controlled at 0.4～0.65 (mass/mass), and 0.15～0.45 second residence time, hydrocarbon partial pressure is 0.17～0.23 MPa, 785～875 ℃ of outlet temperatures.Therefore, sample has more than 300 data, comprises the data of 8 rerum naturas, 4 thermal cracking process conditions and 5 kinds of Pyrolyzate yields in each sample, and the part sample is seen table 1.
According to sample data, write the SVMs program.This program has three parts, i.e. database, SVMs training program, predictor.Database adopts the ACCESS of Microsoft, and the SVMs training program adopts the C language, and its principle is the principle that Vapnik proposes SVMs, and the kernel function of SVMs adopts RBF.
Table 1, SVM part training sample data
| Sequence number | Density | Volume average boiling point ℃ | Hydrogen richness weight % | Correlation index | N-alkane weight % | Isoparaffin weight % | Naphthenic hydrocarbon weight % | Aromatic hydrocarbons weight % | Coil outlet temperature ℃ | Boiler tube top hole pressure MPa |
| 1 | 0.7225 | 119.8 | 15.3418 | 9.23 | 41.09 | 22.60 | 32.23 | 3.62 | 800.0 | 0.200 |
| 2 | 0.7225 | 119.8 | 15.3418 | 9.23 | 41.09 | 22.60 | 32.23 | 3.62 | 815.0 | 0.200 |
| 3 | 0.7225 | 119.8 | 15.3418 | 9.23 | 41.09 | 22.60 | 32.23 | 3.62 | 830.0 | 0.200 |
| 4 | 0.7225 | 119.8 | 15.3418 | 9.23 | 41.09 | 22.60 | 32.23 | 3.62 | 845.0 | 0.200 |
| 5 | 0.7225 | 119.8 | 15.3418 | 9.23 | 41.09 | 22.60 | 32.23 | 3.62 | 860.0 | 0.200 |
Continuous table 1
| The second residence time | Water/oil gram/gram | Hydrogen yield weight % | Methane yield weight % | Yield of ethene weight % | Propene yield weight % | Butadiene yield weight % |
| 0.206 | 0.500 | 0.55 | 9.31 | 23.66 | 16.03 | 4.98 |
| 0.201 | 0.500 | 0.63 | 10.94 | 26.95 | 16.22 | 5.54 |
| 0.200 | 0.500 | 0.67 | 12.33 | 29.38 | 16.01 | 5.72 |
| 0.200 | 0.500 | 0.69 | 13.83 | 31.53 | 15.28 | 5.80 |
| 0.200 | 0.500 | 0.76 | 14.64 | 32.88 | 14.18 | 5.32 |
With the sample data separated into two parts: it is training data that first selects 2/3rds data immediately, remaining 1st/3rd, predicted data.In the training data input database; With the SVMs training program it is carried out learning training then; Form supporting vector machine model, and then call predictor, the rerum natura and the thermal cracking process condition of input predicted data; Calculate the thermal cracking yield of five kinds of pyrolysis products, table 2 is seen in part calculated value and experiment value contrast.
The calculated value of table 2. supporting vector machine model and experiment value are relatively
From table 2, can find out; Experiment value and supporting vector machine model predicted value error are basically in 10%; This explanation adopts supporting vector machine model to predict that the Pyrolyzate yield of naphtha is feasible, and the computing time of application supporting vector machine model is very short, and computing time was less than 1.0 seconds on the computing machine of Pentium IV CPU1.0GHz; Can tackle ethylene plant's raw material and change frequent situation, therefore have good application prospects.
Claims (5)
1. Forecasting Methodology based on the Pyrolyzate yield of SVMs; It is characterized in that; With thermal cracking raw material rerum natura, thermal cracking process condition and product yield data is sample; The supporting vector machine model that foundation utilizes regression function to estimate utilizes supporting vector machine model to predict the Pyrolyzate yield of cracking stock again, and said Forecasting Methodology may further comprise the steps:
(1) sets up sample: collect the different cracking stock of rerum natura; To its Physical Property Analysis record; Then cracking stock and thinning agent are fed in the pyrolysis experiment device; Under different technology conditions, carry out heat scission reaction, analyze and calculate Pyrolyzate yield, the data of sample are made up of cracking stock rerum natura, thermal cracking process condition and pyrolysis product yield;
(2) set up supporting vector machine model: according to cracking stock rerum natura, thermal cracking process condition and pyrolysis product yield, coding, said program is made up of database, SVMs training program and predictor; Database adopts the database of open-source standard interface, with the sample data input database, for the SVMs training program provides the learning training sample; The support vector training program adopts the standard C language exploitation, and this program principle is based on the principle of SVMs; The SVMs training program is called open-source standard interface database; Sample data is imported in the array in the SVMs training program; The support vector training program is carried out learning training to sample data, sets up pattern-recognition with the rule to sample, forms supporting vector machine model; Wherein cracking stock rerum natura and thermal cracking process condition are as the input variable of said model, and Pyrolyzate yield is as output variable;
(3) pyrolysis product yield prediction: predictor calls supporting vector machine model, and cracking stock rerum natura and thermal cracking process condition data are input in the input variable of supporting vector machine model, dopes Pyrolyzate yield through calculating.
2. the Forecasting Methodology of the Pyrolyzate yield based on SVMs as claimed in claim 1 is characterized in that said Forecasting Methodology may further comprise the steps:
(1) set up sample data: collect known cracking stock rerum natura, thermal cracking process condition and pyrolysis product yield as sample data, select sample data over half as the training sample data at random, the samples remaining data are as the forecast sample data;
(2) input database: with forming sample database in the training sample data input database, cracking stock rerum natura and thermal cracking process condition are as input variable, and the pyrolysis product yield is as output variable;
(3) set up supporting vector machine model: set up SVMs training program based on kernel function; Input variable and output variable form mapping relations through supporting vector machine model; Utilize the SVMs training program that the training sample data are carried out learning training, obtain N support vector X through study and training
i *, i=0,1 ..., N, thus supporting vector machine model formed:
Wherein,
The expression support vector, Y
iThe product yield of expression support vector, α
iThe coefficient of representing i support vector, X are the rerum natura and the thermal cracking process conditions of input, and Y (X) representes yield, and K () is the kernel function of SVMs, are selected from a kind of in Gaussian function, polynomial function, lienar for function and the RBF.
(4) set up the supporting vector machine model information database: with parameter information, support vector information and each coefficient storage of the supporting vector machine model of setting up in said database; Parameter information comprises radially basic kernel function type, gamma parameter and support vector number, and support vector information comprises the numerical value of each dimension of support vector;
(5) prediction pyrolysis product yield: in the supporting vector machine model that the raw material rerum natura in the forecast sample data and thermal cracking process condition entry are set up, calculate predicting the outcome of pyrolysis product yield through supporting vector machine model;
(6) confirm supporting vector machine model: what comparison step (5) obtained predicts the outcome and experimental result; When surpassing acceptable scope with the deviation of experimental result if predict the outcome; Change the SVMs kernel function or SVMs kernel function parameter values regulated, and then train again and predict, until the deviation of predicted value and experimental result less than 10%; Determine supporting vector machine model like this, utilize determined this supporting vector machine model prediction cracking stock Pyrolyzate yield.
3. Forecasting Methodology as claimed in claim 1; It is characterized in that; Said thermal cracking raw material rerum natura is density, grace formula distillation data, cracking stock group composition, hydrogen richness or carbon content or ratio of carbon-hydrogen, carboloy residue, molecular weight, correlation index and refraction index, and perhaps said thermal cracking raw material rerum natura is the composition of cracking stock.
4. Forecasting Methodology as claimed in claim 1 is characterized in that, said thermal cracking process condition comprises throws oil mass, boiler tube temperature in, coil outlet temperature, boiler tube inlet pressure, boiler tube top hole pressure, dilution ratio and the residence time.
5. Forecasting Methodology as claimed in claim 1; It is characterized in that said product yield is the yield of hydrogen, methane, ethane, ethene, acetylene, propane, propylene, propine, butane, butylene, butadiene, benzene,toluene,xylene, ethylbenzene, pyrolysis gasoline, pyrolysis gas oil and the fuel oil of cracking generation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810225363XA CN101727609B (en) | 2008-10-31 | 2008-10-31 | Pyrolyzate yield forecasting method based on support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810225363XA CN101727609B (en) | 2008-10-31 | 2008-10-31 | Pyrolyzate yield forecasting method based on support vector machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101727609A CN101727609A (en) | 2010-06-09 |
| CN101727609B true CN101727609B (en) | 2012-02-29 |
Family
ID=42448475
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810225363XA Active CN101727609B (en) | 2008-10-31 | 2008-10-31 | Pyrolyzate yield forecasting method based on support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101727609B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103524284B (en) * | 2013-10-14 | 2015-05-20 | 中国石油化工股份有限公司 | Forecasting and optimizing method for ethylene cracking material configuration |
| CN104765347B (en) * | 2015-03-26 | 2019-03-01 | 华东理工大学 | Yield real-time predicting method in a kind of residual oil delayed coking |
| CN111382489A (en) * | 2018-12-28 | 2020-07-07 | 中国石油天然气股份有限公司 | Method and device for optimizing technological parameters of cracking furnace and cracking furnace |
| CN111598305B (en) * | 2020-04-21 | 2023-07-18 | 汉谷云智(武汉)科技有限公司 | Light hydrocarbon separation device running state optimization and prediction method |
| CN111679636B (en) * | 2020-05-11 | 2021-11-09 | 杭州睿疆科技有限公司 | System, method and computer equipment for processing production process parameters |
| CN113035287B (en) * | 2021-03-09 | 2022-11-04 | 清华大学 | Prediction method for steam thermal cracking process |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1687428A (en) * | 2005-03-24 | 2005-10-26 | 上海交通大学 | Method of soft predicting state variables of biofermentation process based on supporting vector machine |
-
2008
- 2008-10-31 CN CN200810225363XA patent/CN101727609B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1687428A (en) * | 2005-03-24 | 2005-10-26 | 上海交通大学 | Method of soft predicting state variables of biofermentation process based on supporting vector machine |
Non-Patent Citations (2)
| Title |
|---|
| 范红波.基于SVM的柴油机油液光谱预测模型研究.《润滑与密封》.2006,(第11期),全文. * |
| 黄训诚.基于支持向量机在线学习方法的短期负荷预测.《西安交通大学学报》.2005,第39卷(第4期),413页-415页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101727609A (en) | 2010-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102213949B (en) | Method for optimizing value of ethylene device | |
| CN101727609B (en) | Pyrolyzate yield forecasting method based on support vector machine | |
| Gutiérrez-Antonio et al. | Simultaneous energy integration and intensification of the hydrotreating process to produce biojet fuel from Jatropha curcas | |
| CN103524284B (en) | Forecasting and optimizing method for ethylene cracking material configuration | |
| Yang et al. | A hybrid deep learning and mechanistic kinetics model for the prediction of fluid catalytic cracking performance | |
| CN104789256A (en) | Real-time yield predicting method for catalytic cracking device | |
| Wang et al. | Energy and production efficiency optimization of an ethylene plant considering process operation and structure | |
| CN104804761A (en) | Real-time yield prediction method for hydrocracking device | |
| CN106709654A (en) | Global operating condition evaluating and quality tracing method for hydrocracking process | |
| CN103087750B (en) | Naphtha industrial cracking furnace value maximization model construction method | |
| Basak et al. | Industrial hydrocracker model based on novel continuum lumping approach for optimization in petroleum refinery | |
| CN104484714A (en) | Real-time prediction method for catalytic reforming device | |
| CN104361153A (en) | Method for predicting coking amount of heavy oil catalytic cracking settler | |
| CN111475957B (en) | Oil refining process production plan optimization method based on device mechanism | |
| Zhou et al. | Crude oil hierarchical catalytic cracking for maximizing chemicals production: Pilot-scale test, process optimization strategy, techno-economic-society-environment assessment | |
| CN108446358B (en) | Optimization method based on MIV and association rule and data modeling method of petrochemical equipment | |
| Gong et al. | An energy efficiency integration optimization scheme for ethylene production with respect to multiple working conditions | |
| Zhang et al. | Prediction of research octane number loss and sulfur content in gasoline refining using machine learning | |
| CN102289198A (en) | Automatic online control method for production operation of industrial cracking furnace | |
| CN108287474B (en) | Robust operation optimization method of catalytic reforming reactor based on uncertainty of raw materials | |
| CN103087751B (en) | A kind of construction method of industrial pyrolysis furnace value maximization model | |
| CN102289199B (en) | Automatic on-line control method for production and operation of industrial cracking furnace | |
| CN103087753B (en) | Heavy raw material industrial cracking furnace value maximization model construction method | |
| CN103087752B (en) | Industrial cracking furnace cracking product value maximization model construction method | |
| CN115938502A (en) | Chemical product characteristic prediction method and system based on molecular-level reaction mechanism |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |