CN108287474A - Based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material - Google Patents
Based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material Download PDFInfo
- Publication number
- CN108287474A CN108287474A CN201711444462.2A CN201711444462A CN108287474A CN 108287474 A CN108287474 A CN 108287474A CN 201711444462 A CN201711444462 A CN 201711444462A CN 108287474 A CN108287474 A CN 108287474A
- Authority
- CN
- China
- Prior art keywords
- model
- raw material
- reactor
- data
- robust operation
- 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
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
- Production Of Liquid Hydrocarbon Mixture For Refining Petroleum (AREA)
Abstract
The present invention relates to one kind being based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material, includes the following steps：S1, the reactor model for building catalytic reforming；S2, acquisition initial data are simultaneously handled, and generate raw material uncertain data library；S3, the data based on raw material uncertain data library carry out sampling analysis, obtain raw material uncertain data sampling analysis result；S4, it is based on step S1 and step S3, selection needs the operating condition optimized and productive target statistic respectively as optimized variable and optimization aim, builds robust operation Optimized model；S5, solution being optimized to robust operation Optimized model, obtaining Pareto optimal solution sets, which corresponds to an optimal operating condition collection to be selected；S6, it is concentrated from optimal operating condition to be selected according to application demand and chooses intended operating conditions.Compared with prior art, the present invention has many advantages, such as to promote product index and operates robustness, reduces model complexity.
Description
Technical field
The present invention relates to petrochemical industries, and the probabilistic catalytic reforming reactor of raw material is based on more particularly, to one kind
Robust operation optimization method.
Background technology
Catalytic reforming process is one of the main processes in petroleum refining and petrochemical process.It is centainly to operate
Under the conditions of, naphtha is changed into aromatic hydrocarbons (including benzene,toluene,xylene etc.) or high-octane premium, and by-product
The process of hydrogen.Wherein, high-octane premium has good shock resistance, may be used as aviation gasoline and automobile-used vapour
Oil；Aromatic hydrocarbons is important industrial chemicals, can be used as the raw material of coating, dyestuff and pesticide etc.；The hydrogen of by-product is that petrochemical industry adds
The important cheap hydrogen source of hydrogen unit.So catalytic reforming is in petrochemical process, or even entire chemical industry suffers from
Significance.
Catalytic reforming reaction part is the core operation unit in entire technical process, and product form is complicated and various,
The selection of operating condition directly affects the composition, yield and economic benefit of product.In order to improve product yield and economic benefit,
It needs that operating condition is in optimized selection.
It is mainly currently based on the operation optimization method of catalytic reformer system：Under the premise of fixed product form, it is based on
The catalytic reforming mechanism model built in advance, selection need the operating condition that optimizes and productive target respectively as optimized variable and
Optimization aim, with various colony intelligence optimization algorithms (such as genetic algorithm, evolution difference algorithm, artificial bee colony algorithm), to mould
Type optimizes.This optimisation strategy can select optimal operating condition, significantly improve productive target and economic benefit.But
This operation optimization method there is a problem in that：
(1) optimization of operating condition is carried out under the premise of raw material composition is fixed, does not consider the influence of component fluctuation.And show
Field raw material and product form analysis result are shown, are influenced by upstream oil refining apparatus, the raw material composition fluctuation of catalytic reforming process
Obviously, and cause product quality fluctuation apparent.Fixed optimization of operating condition result is formed not necessarily in other originals based on raw material
It is showed under material strip part good.Therefore need during operation optimization, the uncertainty for considering raw material is needed, device operation is improved
Robustness.
(2) operation optimization based on mechanism model is computationally intensive, especially is implementing to consider probabilistic robust operation
When optimization, the computing capability of the far super common computer of calculation amount is not used to real-time optimization.
Invention content
The present invention is exactly to provide a kind of optimal operating condition collection to be selected to overcome the problems of the above-mentioned prior art.
An object of the present invention is for raw material uncertain problem existing for catalytic reforming process, it is proposed that using not
The method of analysis is determined to indicate uncertain problem existing for process, and during operation optimization, considers the not true of raw material
It is qualitative, improve the robustness of device operation.
It is excessive the second object of the present invention is to be directed to operand existing for robust operation optimization process, it cannot achieve dynamic in real time
The defect of state optimization, it is proposed that the side of the mechanism model of original calculation amount costliness is replaced with the agent model based on data-driven
Calculating cost is greatly reduced under the premise of ensureing model accuracy in case.
The purpose of the present invention can be achieved through the following technical solutions：
One kind being based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material, includes the following steps：
S1, structure catalytic reforming reactor mechanism model；
S2, acquisition initial data are simultaneously handled, and generate raw material uncertain data library；
S3, the data based on raw material uncertain data library carry out sampling analysis, obtain the sampling of raw material uncertain data
Analysis result；
S4, be based on step S1 and step S3, selection need the operating condition that optimizes and productive target statistic respectively as
Optimized variable and optimization aim build robust operation Optimized model；
S5, solution being optimized to the robust operation Optimized model, obtaining Pareto optimal solution sets, the Pareto is optimal
Disaggregation corresponds to an optimal operating condition collection to be selected；
S6, it is concentrated from the optimal operating condition to be selected according to application demand and chooses intended operating conditions.
Preferably, the step S1 includes：
S101, the lumped watershed hydrologic model model that catalytic reforming process is established using the method for dividing lump；
S102, reactor model of the structure based on multi-region parallel connection plug flow model.
Preferably, the reactor model based on multi-region parallel connection plug flow model has the property that：
1a, entire reactor include the radially moving bed reactor of four sections of eclipsed forms, and every section of reactor is divided into N in an axial direction
A section；
Each reaction interval in 1b, every section of reactor is regarded as one-dimensional plug flow model radially, and every section anti-
Answer concentration of component, the speed of the inlet of device consistent with temperature；
1c, the last period reactor to rear first stage reactor concentration of component and temperature change meet perfect gas hybrid regulatory
Then；
The reason of 1d, catalyst inactivation includes that acidic site inactivation and metallic site inactivate two kinds；
1e, meet mass conservation law and law of conservation of energy in each reaction interval.
Preferably, in the step S2, initial data is acquired from DCS database and LIMS databases, and according to LIMS numbers
According to the acquisition time of data in library, the related data in DCS database under the corresponding time is extracted.
Preferably, in the step S2, the processing of initial data includes：To the hash and wrong data of initial data
It is rejected, and is recalculated according to constructed reactor model, generate raw material uncertain data library.
Preferably, the principle of the rejecting includes：
The data unrelated with lumped component and product index are rejected in 2a, LIMS database；
Label is micro in 2b, LIMS database " or "<0.1% " data are accordingly to be regarded as 0；
The data that 2c, LIMS database and DCS database repeat record take last time record result；
2d, giving up LIMS databases and DCS database, there are the infull data sets of data record.
Preferably, in the step S3, sampled using the super latin cube methods of sampling, hits when sampling according to
The Sampling uniformity of major product index is preferably obtained.
Preferably, in the step S4, the robust operation Optimized model of structure includes average value standard deviation model, Maximax
Chance-constrained Model or Minimax Chance-constrained Models.
Preferably, in the step S5, optimizing solution to the robust operation Optimized model is specially：
The reactor mechanism model is replaced using the agent model of data-driven, with multi-objective genetic algorithm to described
Robust operation Optimized model optimizes solution, obtains Pareto optimal solution sets.
Preferably, in the step S6, intended operating conditions obtain intended operating conditions by LD decision-making techniques.
Compared with prior art, the invention has the advantages that：
1) present invention is the robust operation optimisation strategy based on uncertainty analysis, improves product index and the robust of operation
Property.
2) it constructs based on probabilistic Multi-objective Robust optimization of operating condition and decision strategy is operated, improves target
The accuracy that operating condition is chosen.
3) present invention proposes the side replaced to original mechanism model using the method for the agent model of data-driven
Case significantly reduces model complexity while ensureing model accuracy, makes it possible to realize that the rolling dynamic of operating condition is excellent
Change.
4) present invention obtains last multiple-objection optimization operating condition using LD decision-making techniques, and it is high as a result to choose accuracy.
Description of the drawings
Fig. 1 is the flow diagram of the method for the present invention；
Fig. 2 is a kind of typical continuous catalytic reforming reaction unit schematic diagram；
Fig. 3 is the 27 lumped watershed hydrologic model model reaction networks established in embodiment；
Fig. 4 is the catalytic reforming reactor model based on multi-region parallel connection plug flow model；
The situation of change schematic diagram of different product index average when Fig. 5 is different hits, wherein (5a) receives for aromatic hydrocarbons
Rate, (5b) are hydrogen yield, and (5c) is heavy aromatics yield, and (5d) is energy consumption；
Fig. 6 is the results contrast schematic diagram in the optimal forward positions Pareto and industrial practical operation condition that optimization obtains；
Fig. 7 is the European norm result figure of the optimization aim in the forward positions corresponding Pareto, wherein (7a) is aromatics yield,
(7b) is energy consumption.
Specific implementation mode
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention
Premised on implemented, give detailed embodiment and specific operating process, but protection scope of the present invention is not limited to
Following embodiments.
As shown in Figure 1, the present embodiment provides one kind based on the probabilistic catalytic reforming reactor robust operation of raw material it is excellent
Change method, includes the following steps：
S1, structure catalytic reforming reactor mechanism model；
S2, acquisition initial data are simultaneously handled, and generate raw material uncertain data library；
S3, the data based on raw material uncertain data library carry out sampling analysis, obtain the sampling of raw material uncertain data
Analysis result；
S4, be based on step S1 and step S3, selection need the operating condition that optimizes and productive target statistic respectively as
Optimized variable and optimization aim build robust operation Optimized model；
S5, solution being optimized to the robust operation Optimized model, obtaining Pareto optimal solution sets, the Pareto is optimal
Disaggregation corresponds to an optimal operating condition collection to be selected；
S6, it is concentrated from the optimal operating condition to be selected according to application demand and chooses intended operating conditions.
Above-mentioned steps are specifically described below, and with a kind of typical continuous catalytic reforming reaction dress shown in Fig. 2
It is set to example.
1, catalytic reforming reactor mechanism model is built
Catalytic reforming reactor mechanism model includes mainly reaction Kinetics Model and reactor model two parts.
1) reaction Kinetics Model
Since catalytic reforming process reactant and reaction network are complicated, each reactant and each reaction can not be built
Mould is studied, therefore the method for dividing lump is used to solve the problems, such as kinetics complicated in catalytic reforming process.So-called lump side
Method exactly calculates physico-chemical property in reaction system and the similar component of kinetic property as a virtual component.
The complex of the present embodiment is to be the catalytic reforming process of main productive target with aromatic hydrocarbons, therefore have chosen aromatic hydrocarbons
27 lumped watershed hydrologic model network struction reaction Kinetics Models of component subdivision.Its reaction network figure is shown in Fig. 3.
2) reactor model
The distributed intelligence of the lumped component and temperature in catalytic reforming reactor is obtained in order to be more accurate, and considers to urge
The influence of agent inactivation, the present embodiment use the reactor model based on multi-region parallel connection plug flow model, schematic diagram such as Fig. 4
It is shown.The reactor model feature based on multi-region parallel connection plug flow model is as follows：
A. entire reactor assembly includes the radially moving bed reactor of four sections of eclipsed forms, and every section of reactor divides in an axial direction
At N number of section；
B. each reaction interval in every section of reactor is regarded as one-dimensional plug flow model radially, and every section of reaction
The concentration of component of the inlet of device, speed, temperature are consistent；
C. the concentration of component and temperature change of the last period reactor to rear first stage reactor meet perfect gas mixing rule,
Formula is as follows：
Wherein,Represent the inlet component concentration of i-th of lumped component of m+1 sections of reactors；Represent m sections
The outlet component concentration of i-th of lumped component of reactor；Represent in m sections of reactors n-th axial subregion last
The outlet component concentration of i-th of lumped component in a radial direction subregion；Represent m sections of reactors outlet component concentration and
Temperature；Represent in m sections of reactors the outlet component temperature in n-th of axial subregion the last one radial subregion.
D. include two kinds of acidic site inactivation and metallic site inactivation the reason of catalyst inactivation, activity is in axial direction
Variation formula it is as follows：
Wherein, h is the axial position of reactor；ametalAnd bmetalIt is the mistake of metal active site in the axial direction respectively
The factor living；aacidAnd bacidIt is the inactivation factor of acidic active sites in the axial direction respectively；WithRespectively represent catalysis
Vivacity of the agent in metallic site and acidic site.
E. meet mass conservation law and law of conservation of energy in each section, formula is as follows：
Mass conservation law：
Law of conservation of energy：
Wherein,Represent the concentration of component of i-th of lumped component in m sections of reactors in n-th of axial subregion, T(nm)The component temperature in n-th of axial subregion in m sections of reactors is represented, z is the radical length of reactor, zmIt is m sections anti-
Answer the radical length of device；QnmIt is the flow in m sections of reactors in n-th of section；ΔrHjIt is the reaction enthalpy of j-th of reaction；
cp,iIt is the specific heat capacity of i-th of lumped component；hnIt is the axial position of n-th of axial subregion, rjIt is the reaction speed of j-th of reaction
Rate.
2, raw data acquisition and processing
1) raw data acquisition
The main source of initial data includes DCS (live distribution type control system) databases and LIMS (laboratory informations
Management system) database two parts.Wherein include from the main information of DCS database acquisition：Feed loading, cycle hydrogen amount, hydrogen
The gas amount of sending outside, reactor and follow-up rectifying column rate of discharge, reactor inlet temperatures at different levels, each stage reactor temperature drop, operation pressure
Power, reactor pressure decrease at different levels etc..From LIMS databases acquire main information include：Reaction feed oil, sends hydrogen outside at recycle hydrogen
The group that gas, reaction generate oil etc. is grouped as situation.The data volume of general LIMS databases is much smaller than DCS database, therefore according to
The acquisition time of data in LIMS databases, corresponds to the related data in the DCS database of time in extraction.So far, it completes
Whole raw material data collection tasks.
2) original data processing
Due to industrial field data complexity, especially component analysis data more by human interference, it is understood that there may be leakage
Note, incorrect posting, the case where remembering again, therefore need to reject the useless and wrong data of initial data.The main principle of rejecting is such as
Under：
The data unrelated with lumped component and product index are rejected in a.LIMS databases；
Label is micro in b.LIMS databases " or "<0.1% " data are accordingly to be regarded as 0；
The data that c.LIMS databases and DCS database repeat record only take last time record result；
D.LIMS databases and DCS database have that data record is not complete, then give up this data set, be not included in original
Beginning database data.
LIMS databases and DCS database corresponding data after processing is as initial data base.
3) raw material aggregate data calculates
Since the lump division methods of initial data base record are not necessarily corresponding with selected lumped model, therefore need to press
The lumped component data in initial data base are recalculated according to 27 lumped watershed hydrologic model model partition methods of selection,
And the flow of association reaction feed oil and recycle hydrogen, it calculates different data and concentrates, the flow rate of each lumped component in raw material, and import
Raw material uncertain data library.
3, raw material uncertain data sampling analysis
Taking and spend larger due to the test of oil product component, test frequency is relatively low, can obtain raw material uncertain data
Measure on the low side, and covering surface is little, therefore needs to carry out resampling according to current uncertain data set pair component data, to improve
Data accuracy.
The present embodiment is directed to 30 groups of raw material uncertain datas, calculates the fluctuation bound of each lumped component flow rate.And with this
Bound is sampled as it, is sampled using the super latin cube methods of sampling, to ensure the uniformity of its sampling.
The selection of hits can influence sampling precision and operation time.Generally, hits is more, and sampling precision is higher, but
Corresponding operation time is also longer.Therefore when choosing different hits, according to the variation of the mean value of major product index and standard deviation
The preferred hits of situation.The situation of change of different product index average when Fig. 5 is different hits, according to as a result, the present embodiment
In preferred 1000 hits.
4, robust operation Optimized model is built
The type of robust operation Optimized model can be chosen according to actual needs, including：
1) average value standard deviation model
Wherein, fi(x) it is the different optimization aims chosen, the optimization aim used is recommended to have：Aromatics yield YAMaximum,
Light aromatics yield YLAMaximum, hydrogen yield YHMaximum, heavy arene yield YHAMinimum, light paraffins yield YLPMinimum, four
Heat exchanger energy consumption EC minimums etc..X is optimized variable, includes the inlet temperature (T of four sections of reactors1, T2, T3, T4), hydrogen-oil ratio
(HC) and operating pressure (P)；μ is the mean value of optimization aim；σ is the standard deviation of optimization aim；XlbAnd XubRespectively optimized variable
Bound.
2) Maximax Chance-constrained Models
Wherein, fi(x) it is the different optimization aims chosen, αiFor the confidence level of selection, fiOptimize under confidence level to formulate
Mesh target value, ε are a random vector
3) Minimax Chance-constrained Models
Minimax Chance-constrained Models are used in the present embodiment, with aromatics yield YAMaximum and four heat exchanger energy consumptions
The minimum optimization aims of EC, confidence level are selected as 5%, and the function expression of Model for Multi-Objective Optimization is in the present embodiment：
Wherein, Xlb=[788,788,788,788,1.5,0.5], Xub=[808,808,808,808,2.5,0.55].
5, the optimisation strategy of the agent model based on data-driven
Based on above-mentioned robust operation optimization problem and catalytic reforming reactor model, it is based on using multi-objective genetic algorithm,
Iteration optimization, the optimal forward position disaggregation of Pareto to obtain optimization aim are executed to model, corresponding operating condition is to wait for
The optimal operating condition of choosing.But this method computation complexity is very high, it is contemplated that the deadline (was based on 12 cores up to 5000 hours
CPU calculates).Computational efficiency is improved while in order to ensure computational accuracy, the present invention is replaced using the agent model of data-driven
After the larger reactor mechanism model of calculation amount, then with multi-objective genetic algorithm solution is optimized, effectively reduces optimization
Complexity.
The reactor mechanism model is replaced using the agent model of data-driven, with multi-objective genetic algorithm to described
Robust operation Optimized model optimizes solution, obtains Pareto optimal solution sets.
In the present embodiment, the method for two different agent models based on data-driven is taken to replace original calculation amount
Huge mechanism model.
(1) optimization method based on offline agent model, step are：
A. the bound based on optimized variable, using Latin Hypercube Sampling method acquisition N groups optimized variable vector.
B. N group optimized variables are brought into mechanism model and solves corresponding object function statistics figureofmerit.
C. using optimized variable as input, object function counts figureofmerit as output, utilizes kriging model construction generations
Model is managed, corresponding model parameter is obtained.
D. based on the agent model newly created, with multi-objective genetic algorithm, iteration optimization is executed to model, to obtain
The optimal forward position disaggregation of Pareto of optimization aim.
The calculating time of this method is down to about 24 hours, and efficiency is improved more than 200 times.The optimal behaviour obtained by this method
Former mechanism model is brought into as condition and verifies its model error, and the mean absolute error of two of which index is respectively 0.012% He
0.058GJ/h, in the reasonable scope.
(2) optimization method is acted on behalf of online, and step is：
A. the bound based on optimized variable, using Latin Hypercube Sampling method acquisition M groups optimized variable vector.
B. M group optimized variables are brought into mechanism model and solves corresponding object function statistics figureofmerit.
C. using optimized variable as input, object function counts figureofmerit as output, and mould is acted on behalf of using RBF model constructions
Type obtains corresponding model parameter.
D. the agent model based on acquisition obtains its forward positions Pareto point, then verifies it with multi-objective genetic algorithm
Whether it is Pareto optimal, preserves noninferior solution, the information of the agent model of structure is used in combination to go to select the optimization of next group selection to become
Amount vector.
E. step b to step d is carried out using the optimized variable vector circulant newly selected, until end condition is reached.
F. the optimal forward position disaggregation of Pareto of the optimization aim finally obtained, a corresponding optimal operating condition collection to be selected.
The calculating time of this method is down to about 24 hours, and efficiency is improved more than 200 times, and model predication value is that model is true
Real value.
The optimal forward positions Pareto of above two method acquisition and industrial practical operation condition is respectively adopted in the present embodiment
Results contrast is as shown in Figure 6.Compared with current operating condition, aromatics yield rises about 0.80%, and energy consumption declines about 13.21GJ/
h。
6, intended operating conditions are chosen by decision-making technique
Specifically, the present embodiment obtains last multiple-objection optimization operation using LD (Level diagrams) decision-making technique
Condition.Specific steps in decision-making is as follows：
A. in the optimal forward positions m optimization aim Pareto of acquisition, solution is concentrated, find out each optimization aim minimum value and
Maximum value.Formula is as follows：
Wherein, X*It is the optimal forward position disaggregation of Pareto that multiple-objection optimization obtains；WithIt is m-th of optimization respectively
The maximum value and minimum value of target.
B. then according to maximum value and minimum value, optimization aim is normalized, formula is as follows：
Wherein,To normalize result.
If c. Optimized model is minimization problem,It is worth smaller, represents solution and got over respect to m-th of target of the moon
Close to ideal point.Three kinds of normal forms are used to the degree of closeness between assessment Pareto forward position disaggregation and ideal point, are respectively：
Single order norm, Euclidean Norm and Infinite Norm, formula difference are as follows：
The present embodiment is calculated using European norm.
D. European Norm minimum is selected, the optimal solution under the decision-making technique and its corresponding object run variable are obtained.This
The object run variable that optimization method of the embodiment based on offline agent model is finally chosen is：
X=[796.03,788.00,788.32,793.05,1.500,0.500]
Corresponding aromatics yield and energy consumption Minimax (5% confidence level) statistics figureofmerits are respectively 63.52% He
174.80GJ/h.It is run 30 days with the operating condition, relatively industrial actual motion condition, aromatics yield mean value are promoted at present
0.85%, energy consumption mean value reduces 8.81GJ/h.Fig. 7 is the European norm result figure of the optimization aim in the corresponding forward positions Pareto.
The present embodiment fluctuates data based on 30 days raw materials of catalytic reforming process, proposes a set of Shandong based on uncertainty analysis
Optimisation strategy is made in clubs, improves the robustness of product index and operation, while proposing the agent model using data-driven
Method scheme that original mechanism model is replaced significantly reduce model complexity, being allowed to can while ensureing model accuracy
To realize the rolling dynamic optimization of operating condition.
The preferred embodiment of the present invention has been described in detail above.It should be appreciated that those skilled in the art without
It needs creative work according to the present invention can conceive and makes many modifications and variations.Therefore, all technologies in the art
Personnel are available by logical analysis, reasoning, or a limited experiment on the basis of existing technology under this invention's idea
Technical solution, all should be in the protection domain being defined in the patent claims.
Claims (10)
1. one kind be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material, which is characterized in that including with
Lower step：
S1, structure catalytic reforming reactor mechanism model；
S2, acquisition initial data are simultaneously handled, and generate raw material uncertain data library；
S3, the data based on raw material uncertain data library carry out sampling analysis, obtain raw material uncertain data sampling analysis
As a result；
S4, it is based on step S1 and step S3, selection needs the operating condition optimized and productive target statistic respectively as optimization
Variable and optimization aim build robust operation Optimized model；
S5, solution is optimized to the robust operation Optimized model, obtain Pareto optimal solution sets, the Pareto optimal solution sets
A corresponding optimal operating condition collection to be selected；
S6, it is concentrated from the optimal operating condition to be selected according to application demand and chooses intended operating conditions.
2. according to claim 1 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, the step S1 includes：
S101, the lumped watershed hydrologic model model that catalytic reforming process is established using the method for dividing lump；
S102, reactor model of the structure based on multi-region parallel connection plug flow model.
3. according to claim 2 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, the reactor model based on multi-region parallel connection plug flow model has the property that：
1a, entire reactor include the radially moving bed reactor of four sections of eclipsed forms, and every section of reactor is divided into N number of area in an axial direction
Between；
Each reaction interval in 1b, every section of reactor is regarded as one-dimensional plug flow model radially, and every section of reactor
The concentration of component of inlet, speed it is consistent with temperature；
1c, the last period reactor to rear first stage reactor concentration of component and temperature change meet perfect gas mixing rule；
The reason of 1d, catalyst inactivation includes that acidic site inactivation and metallic site inactivate two kinds；
1e, meet mass conservation law and law of conservation of energy in each reaction interval.
4. according to claim 1 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, in the step S2, acquires initial data from DCS database and LIMS databases, and according in LIMS databases
The acquisition time of data extracts the related data in DCS database under the corresponding time.
5. according to claim 4 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, in the step S2, the processing of initial data includes：The hash and wrong data of initial data are picked
It removes, and is recalculated according to constructed reactor model, generate raw material uncertain data library.
6. according to claim 5 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, the principle of the rejecting includes：
The data unrelated with lumped component and product index are rejected in 2a, LIMS database；
Label is micro in 2b, LIMS database " or "<0.1% " data are accordingly to be regarded as 0；
The data that 2c, LIMS database and DCS database repeat record take last time record result；
2d, giving up LIMS databases and DCS database, there are the infull data sets of data record.
7. according to claim 1 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, in the step S3, is sampled using the super latin cube methods of sampling, hits when sampling is according to main production
The Sampling uniformity of product index is preferably obtained.
8. according to claim 1 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
Be characterized in that, in the step S4, the robust operation Optimized model of structure include average value standard deviation model, Maximax chances about
Beam model or Minimax Chance-constrained Models.
9. according to claim 1 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, in the step S5, optimizing solution to the robust operation Optimized model is specially：
The reactor mechanism model is replaced using the agent model of data-driven, with multi-objective genetic algorithm to the robust
Operation optimization model optimizes solution, obtains Pareto optimal solution sets.
10. according to claim 1 be based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material,
It is characterized in that, in the step S6, intended operating conditions obtain intended operating conditions by LD decision-making techniques.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711444462.2A CN108287474B (en) | 2017-12-27 | 2017-12-27 | Robust operation optimization method of catalytic reforming reactor based on uncertainty of raw materials |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711444462.2A CN108287474B (en) | 2017-12-27 | 2017-12-27 | Robust operation optimization method of catalytic reforming reactor based on uncertainty of raw materials |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108287474A true CN108287474A (en) | 2018-07-17 |
| CN108287474B CN108287474B (en) | 2021-11-09 |
Family
ID=62832421
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711444462.2A Active CN108287474B (en) | 2017-12-27 | 2017-12-27 | Robust operation optimization method of catalytic reforming reactor based on uncertainty of raw materials |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108287474B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110187635A (en) * | 2019-04-10 | 2019-08-30 | 浙江中控软件技术有限公司 | Real-time optimization method and apparatus for continuous reformer |
| CN112782979A (en) * | 2020-12-25 | 2021-05-11 | 杭州电子科技大学 | Real-time optimization control system and method for continuous catalytic reforming device |
| CN114970375A (en) * | 2022-07-29 | 2022-08-30 | 山东飞扬化工有限公司 | Rectification process monitoring method based on real-time sampling data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289579A (en) * | 2011-08-03 | 2011-12-21 | 浙江大学 | Modeling method for 38-lumping continuous reforming device reactor |
| CN102867083A (en) * | 2012-08-30 | 2013-01-09 | 浙江大学 | High-rigidity and light-weight design method considering uncertainty of slide block mechanism of press machine |
| CN104484714A (en) * | 2014-11-26 | 2015-04-01 | 华东理工大学 | Real-time prediction method for catalytic reforming device |
-
2017
- 2017-12-27 CN CN201711444462.2A patent/CN108287474B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102289579A (en) * | 2011-08-03 | 2011-12-21 | 浙江大学 | Modeling method for 38-lumping continuous reforming device reactor |
| CN102867083A (en) * | 2012-08-30 | 2013-01-09 | 浙江大学 | High-rigidity and light-weight design method considering uncertainty of slide block mechanism of press machine |
| CN104484714A (en) * | 2014-11-26 | 2015-04-01 | 华东理工大学 | Real-time prediction method for catalytic reforming device |
Non-Patent Citations (7)
| Title |
|---|
| X-J.DONG 等: "Multi-Zone Parallel-Series Plug Flow Reactor Model with Catalyst Deactivation Effect for Continuous Catalytic Reforming Process", 《CHEMICAL ENGINEERING SCIENCE》 * |
| YANG KE 等: "Multi-objective steady-state optimization of two-chamber microbial fuel cells", 《CHINESE JOURNAL OF CHEMICAL ENGINEERING》 * |
| 侯卫锋: "催化重整流程模拟与优化技术及其应用研究", 《中国博士学位论文全文数据库 工程科技Ⅰ辑》 * |
| 周红军: "芳烃型连续重整集总反应动力学模型研究", 《中国博士学位论文全文数据库 工程科技Ⅰ辑》 * |
| 孙守峰 等: "催化重整固定床反应器反应条件的优化", 《石油学报（石油加工）》 * |
| 李鸿亮 等: "基于混合遗传算法的催化重整过程多目标优化", 《化工学报》 * |
| 江洪波 等: "集总动力学模型研究的普遍方法与原则探讨", 《华东理工大学学报》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110187635A (en) * | 2019-04-10 | 2019-08-30 | 浙江中控软件技术有限公司 | Real-time optimization method and apparatus for continuous reformer |
| CN110187635B (en) * | 2019-04-10 | 2022-08-02 | 浙江中控软件技术有限公司 | Method and apparatus for real-time optimization of continuous reformers |
| CN112782979A (en) * | 2020-12-25 | 2021-05-11 | 杭州电子科技大学 | Real-time optimization control system and method for continuous catalytic reforming device |
| CN114970375A (en) * | 2022-07-29 | 2022-08-30 | 山东飞扬化工有限公司 | Rectification process monitoring method based on real-time sampling data |
| CN114970375B (en) * | 2022-07-29 | 2022-11-04 | 山东飞扬化工有限公司 | Rectification process monitoring method based on real-time sampling data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108287474B (en) | 2021-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Babazadeh et al. | A sustainable second-generation biodiesel supply chain network design problem under risk | |
| Han et al. | Production capacity analysis and energy saving of complex chemical processes using LSTM based on attention mechanism | |
| Arriola et al. | A bibliometric review on the application of fuzzy optimization to sustainable energy technologies | |
| Terrazas‐Moreno et al. | Simultaneous design, scheduling, and optimal control of a methyl‐methacrylate continuous polymerization reactor | |
| CN108287474A (en) | Based on the probabilistic catalytic reforming reactor robust operation optimization method of raw material | |
| Deng et al. | Integrating feature optimization using a dynamic convolutional neural network for chemical process supervised fault classification | |
| Yin et al. | An integrated computational intelligence technique based operating parameters optimization scheme for quality improvement oriented process-manufacturing system | |
| Wang et al. | A deep learning-based robust optimization approach for refinery planning under uncertainty | |
| Cheng et al. | Scheduling flexible manufacturing cell with no-idle flow-lines and job-shop via Q-learning-based genetic algorithm | |
| Jove et al. | Modeling of bicomponent mixing system used in the manufacture of wind generator blades | |
| Yuan et al. | Spatiotemporal attention mechanism-based deep network for critical parameters prediction in chemical process | |
| Li et al. | Heat exchanger network synthesis considering detailed thermal-hydraulic performance: Methods and perspectives | |
| Long et al. | Diesel blending under property uncertainty: A data-driven robust optimization approach | |
| Venkata Rao et al. | Constrained economic optimization of shell-and-tube heat exchangers using a self-adaptive multipopulation elitist-jaya algorithm | |
| Zhou et al. | Optimization of the ethane thermal cracking furnace based on the integration of reaction network | |
| Lavric | Process Design, Integration and Optimisation: advantages, challenges and drivers | |
| Bahroun et al. | Artificial intelligence applications in project scheduling: A systematic review, bibliometric analysis, and prospects for future research | |
| Rangaiah et al. | Differential evolution in chemical engineering: developments and applications | |
| Kramer et al. | Artificial intelligence in process control applications and energy saving: a review and outlook | |
| Wu et al. | Applications of AI techniques to generation planning and investment | |
| Dang et al. | Daily and seasonal heat usage patterns analysis in heat networks | |
| Jiang et al. | Assembly system configuration design for reconfigurability under uncertain production evolution | |
| Dong et al. | Robust optimal operation of continuous catalytic reforming process under feedstock uncertainty | |
| Duc et al. | A hybrid methodology based on machine learning for a supply chain optimization problem | |
| CN110728031B (en) | Multi-objective optimization method for balancing complex petrochemical process production energy based on ANN modeling |
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 |