CN1632069A - Intelligent blending method for gasoline octane value - Google Patents

Abstract

Disclosed is an intelligent reconciliation method of octanevalue in gasoline. It uses reconciliation information input to receive the information from user; calculate the basic reconciliation dispensation and quality index in accordance with octanevalue, building of parameterized octanevalue module and other information, which is used as the basis for optimizing dispensation further more; provide the quantity of each component joining in reconciliation, with the goal of getting the maximum economic benefit in accordance with each economic index of component oil and product oil, parameterized octanevalue module, constraint condition of resource in factory and demand in market, which is output by visible mode; and by comparing the practical octane value and calculating their difference, and the difference of practical proposal and the on in theory, use matching learning method to do intelligent adjustment to the parameter of module. The invention uses parameterized series of module, do optimization calculation to reconciliation dispensation on the basis of meeting the quality index of each gasoline, for acquiring the maximum interesting rate of reconciliation production. It has wide range of application and convenient use.

Description

The intelligent blending method of gasoline octane rating

Technical field

The present invention relates to petrochemical industry, specifically a kind of intelligent blending method of gasoline octane rating.

Background technology

Gasoline is a kind of many hydrocarbon mixtures of oil through obtaining after refining, and the index of weighing the gasoline attribute is a lot, as: density, boiling point, susceptibility, anti-knocking property and saturated vapor pressure etc.Wherein, using maximum effect factor of gasoline is the anti-knocking property of gasoline.The anti-knocking property of gasoline are relevant with the combustion processes of gasoline, generally represent with octane value.

Because the preheating response situation difference of different hydrocarbons, anti-knocking property are different, so the various different gasoline octane rating difference of forming.The factor that influences gasoline octane rating is a lot, and most important two aspects are:

(1) place of production of oil

Oil is formed through the very long earth history age by being deposited on subterranean ancient biological remains, because the difference of organism, the difference and the difference in earth history age in residing geographical position, caused oil attribute that the different oil minings of different areas even areal go out between all have very big difference.Therefore, also just there is very big difference in the character of the gasoline that obtains through same refining process of these oil.

Processing parameter when (2) refining

In refining oil process, different processing parameters also can produce certain influence to the attribute of the gasoline after refining.The gasoline that directly obtains from refining equipment, because nature difference is very big, so normally can't directly apply in the middle of productions, the life, these gasoline can be referred to as work in-process gasoline (title again: component gasoline).People are employed at ordinary times all to be the little gasoline product of attribute difference.Gasoline product is to be undertaken obtaining behind the thorough mixing by certain standard by different work in-process gasoline.This process is exactly the petroleum products blending process.For gasoline, other attribute except that octane value generally can guarantee by refining process, so the controlling index of blending process mainly is the octane value of gasoline.

Gasoline concoction be with octane value be higher than respectively the target octane value (claim again: the label of gasoline) and two or more component gasoline that is lower than the target octane value carry out thorough mixing, the octane value of mixed gasoline must not be lower than the target octane number requirement.Because the minimum standard of quality control when the target octane value only is blending is so if the gasoline octane rating after the blending far above the target octane value, is qualified from the quality angle.But also its price is high more owing to the high more gasoline of octane value, so from the angle of economic benefit, above situation is underproof.Be not higher than the target octane value and add in 1～2 the scope so need quality index be controlled at according to different requirements usually during actual the blending.For example: when blending 93# gasoline (octane value is 93), the then actual gasoline octane rating of concocting out should be greater than 93, less than 94～95.Otherwise, also off quality.

Vapour, diesel oil are the main production kinds of most refineries, in actual production, has consequence, growth for the business economic benefit has very big influence, so be domestic or petroleum products blending all quite paid attention to, has obtained a lot of scientific payoffss through the research of decades abroad.These achievements in research can be summarized as following two aspects:

The mathematical computations of (1) petroleum products blending quality index

Mainly be that purpose is to provide a basic reference model for obtaining maximum economic benefits under the condition that satisfies the blending quality according to the mathematical relation of blending mechanism research oil property, blending proportioning and petroleum products blending quality index.For example, having aspect the calculating of polycomponent gasoline octane rating: Nelson (nineteen fifty-five) proposes the linear combination model, Scott people's such as (1958) nonlinear model, the polycomponent gasoline octane rating estimation equations that (1959) such as the healy of Ethyl company propose etc., domestic since the eighties in 20th century also have many people to propose various computation models and method of calculation.As (1981) such as Peng Piao studied gasoline octane rating and form between relation, Tong Deshun (1983) is to the research of blending gasoline octane rating calculation formula, (1984) such as Shen shirt pines to the research of MTB improvement gasoline octane rating, Chen Xinzhi (1997) is to concocting gasoline research method octane value Study of model etc.

(2) calculating of blending production scheduling scheme

On the basis of blending quality index mathematical computations, be the calculating that target is concocted the production scheduling scheme to obtain optimum economic benefit.Main method has linear programming, goal programming etc., as Liu Xiantao (1997) STM (Step Method) method is incorporated into gasoline concoction calculating etc.

The present subject matter that exists: existing gasoline octane value model requires to have accurate octane value and interactively to component oil.What have replaces nonlinear model with linear model, and precision is not enough, narrow limits.A model is accurate to this batch data, changes a collection of just inaccurately, makes that the petroleum products blending model can't practical application, the petroleum products blending scheduling scheme that is difficult to obtain.

Summary of the invention

In order to overcome the problem that the single model scope of application is narrow, be not easy to use, the objective of the invention is to according to the machine learning principle, provide that a kind of scope of application is wide, the intelligent blending method of the gasoline octane rating being convenient to use, it adopts multi-model, on the basis of satisfying various quality of gasoline indexs, the blending prescription is optimized calculating, so that obtain the maximum profit rate that blending is produced.

To achieve these goals, technical scheme of the present invention is: comprise that the input of blending information, octane value calculate, concoct formulation optimization calculating, the visual output of scheme, parameter on-line study, the information from the user is accepted in the input of wherein blending information, the index of component oil, the index of processed oil and to the evaluation and the selection of various blending schemes, and deposit information in corresponding database; Described octane value calculate be according to component oil octane value, build parametrization octane value model and associated information calculation goes out basis blending prescription and quality index, and deposit data the octane value database of prediction in, carry out the basis of formulation optimization as next step; It is resource constraint and market demand in the each economic target, parametrization gasoline octane value model, factory according to component oil, processed oil that described blending formulation optimization calculates, with the maximum economic benefit is the quantity (blending prescription) that target provides each component oil of participating in blending, and deposits related data in blending scheduling scheme database; The visual output of described scheme is that the blending scheduling scheme is outputed on the computer screen with visual phraseology with chart, process flow sheet, the layout drawing that installs pipes etc., for dispatcher's reference; Described parameter on-line study is by the error of error, practical solution and the theoretical scheme of actual octane value and calculated octane number relatively, with machine learning method model parameter is carried out intellectuality adjustment, and will adjust the result and deposit the model parameter database in;

Described parametrization gasoline octane value model is the basis of gasoline octane rating intelligence blending, and it comes from following octane value model, and is variable with the component oil ratio:

$R_m=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{X}_i\frac{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{X}_j{R}_{\mathrm{ij}}{Q}_{\mathrm{ij}}}{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{X}_j{Q}_{\mathrm{ij}}}---\left(1\right)$

Wherein:

R _mThe blended gasoline octane value;

X _jThe massfraction of component j;

R _IjComponent i (is generally (R to the octane value of component j _i+ R _j)/2);

Q is not only the function of component oil type, also is the dispersing function of relative proportion between the component oil, i.e. Q simultaneously _Ij=Q (X _Ij, X _Ji), making described is the adjustable parameter model of variable with the component oil ratio;

Described intelligentized on-line study method is used for accurately describing the influence to octane value of blended gasoline octane value and component oil ratio variation relation and component oil type, and parametrization gasoline octane rating Model Calculation method is:

As calculated octane number R _cWith actual octane value R _fGap Δ R=R _f-R _c, in the time of not in allowed band, parameter Q learnt and adjust according to error; The concrete grammar of adjusting is:

Q _i(k+1)＝Q _ij(k)+β*ΔQ _ij(k) (2)

Here,

$\mathrm{\Δ}{Q}_{\mathrm{ij}}=\mathrm{\ΔR}\frac{\∂R}{\∂Q_{\mathrm{ij}}}$

Formula 2 is the formulas that iterate; Wherein k represents to learn number of times; 0≤β≤1 is the study step-length; Δ R=R _f-R _cBe laboratory octane number R _fWith theoretical octane value R _cPoor; Δ Q _IjBe Q _IjIncrement, Be that R is to Q _IjPartial differential, as Δ Q _IjNear zero, Q _IjWhen (k+1) tending towards stability

Study finishes;

Described allowed band is the permissible error limit that 0＜Δ R＜user sets; The idiographic flow that described parametrization gasoline octane rating Model Calculation parameter is used is: calculate the relative proportion between each component oil, judge and check the model parameter Q that whether exists between the component oil in twos through study, there are not then 2 study by formula of model parameter Q, carry out blending octane value again and calculate through study; As have model parameter Q, then directly calculating parameter Q is set by initial value; The idiographic flow of described parametrization gasoline octane rating Model Calculation parameter adjustment is: will newly fill a prescription as current formulation, each component oil type and ratio in current formulation are provided with coefficient of mutual influence, the number of times of coefficient of mutual influence by study sorted from less to more, calculate the octane value of current formulation, allow the prescription octane value difference actual octane value that equals to fill a prescription deduct the formula calculation octane value, judge whether prescription octane value difference meets the demands, do not meet the demands, select one group of coefficient of mutual influence as adjusting object, and according to prescription octane value difference it is adjusted, return back to the octane value that calculates current formulation; Meet the demands, judge again whether all prescriptions all meet the demands, and all do not meet the demands as all prescriptions, return back to the octane value that calculates current formulation after selecting next assembly side as current formulation again; All meet the demands as all prescriptions, then parameter learning, adjustment algorithm finish;

Described gasoline concoction formulation optimization method is undertaken by " loose coupling " petroleum products blending formulation optimization thought: provide one group of blending prescription by experience earlier, prescription is optimized processing, verify according to octane value calculating parameter gasoline octane value model whether this prescription meets index request again, as the then end that meets the requirements is calculated, otherwise calculation result is returned to described gasoline concoction formulation optimization process, the reference or the basis of carrying out next step optimization as computation optimization model (parametrization gasoline octane rating); Specifically can adopt cost optimization method, quantity optimization method and profit optimization method; Described cost optimization method idiographic flow is: earlier component oil is sorted from small to large by octane value, find the dividing point of high-octane number component oil and low octane rating component oil; The component oil that the user is chosen just is divided into two groups according to octane value again, make up respectively in twos, and octane number requirement according to target, the cost that calculates every kind of combination reaches the wherein ratio of various component oils, and with minimum one group of cost as basic optimum combination, never take out a kind of conduct in the component oil in optimizing the basis combination and newly add component oil, from optimize the basis combination, select each a kind of component oil of height as displacement component oil, to newly add component oil replaces by mode such as equal proportion such as cost such as grade with displacement component oil, judge then whether the result after the displacement can carry out cost optimization, as carrying out cost optimization, then increase the new ratio that adds component oil, judge whether existing component oil reaches the ratio restriction in three kinds of component oils, reach the ratio restriction then with the component oil in the current component oil note do optimization basis combination, judge whether again also to exist and do not add the component oil of optimizing the basis combination, be to take out a kind of conduct in the component oil that then is back to never in optimizing the basis combination newly to add component oil, select each a kind of component oil of height as displacement component oil in the combination of basis from optimizing, otherwise the cost optimization algorithm finish; Do not reach the ratio restriction, be back to and replace by mode such as equal proportion such as cost such as grade with displacement component oil newly adding component oil; As carrying out cost optimization, then adjust the ratio of displacement component oil, make basic quality index of compose of panel just reach target call, calculate this cost, whether the cost of doing this then is less than the judgement of the optimization cost of last time, when the optimization cost that is not less than last time up to increasing the new ratio that adds component oil, the ratio of current each component oil of record is as optimized proportion, again to increasing the new ratio that adds component oil when less than the optimization cost of last time;

Described quantity optimization method idiographic flow is: component oil is sorted from small to large by octane value, find the dividing point of high-octane number component oil and low octane rating component oil, note is according to the tank farm stock of each component oil, blending octane value when calculating all components oil is concocted together, the blending result is higher than requirement and then turns to the component of high group is handled, meet the demands to the quantity maximization up to the octane value that calculates, withdraw from the computation optimization program; The blending result is lower than requirement and then turns to the component of low group is handled, and meets the demands up to the octane value that calculates, and to the quantity maximization, withdraws from the computation optimization program; The blending result meets the demands and then directly carries out the quantity maximization, withdraws from the computation optimization program; Described profit optimization method is boundary with the target octane value, choose the user to such an extent that component oil just is divided into two groups according to octane value, again the component oil of height in two groups carried out combinations of pairs in twos, according to target octane value calculates ratio separately, and with tank farm stock separately as usage quantity but (residual content), but then with usage quantity (residual content) separately, calculate every kind of obtainable profit of combination, choose a kind of combination back of profit maximum and judge whether maximum profit is less than or equal to 0, maximum profit is not less than usage quantity and the residual content calculated after this group component oil of choosing is concocted in this ratio at 0 o'clock, the usage quantity of combination is newly added in total usage quantity of prescription, component oil mass to the quality index of adjusting prescription meets the requirements of scope, return back to then with tank farm stock separately as usage quantity but (residual content), but, finished profit as maximum profit less than 0 o'clock and optimize algorithm then with usage quantity (residual content) separately.

Compared with prior art, the present invention has more following advantage:

1. the present invention has overcome that precision when replacing nonlinear model with linear model is not enough, the disadvantage of narrow limits.Avoided existing nonlinear model only accurate by the on-line study method, the problem that applicable surface is narrow to certain class data.

2. owing to adopt parametrization blending model, make calculating convenient, practical.The user can select different formulation optimization processes, makes the more realistic demand of optimum result.

Description of drawings

Fig. 1 is a system architecture of the present invention.

Fig. 2 is the relation between computation optimization process and the octane value model (Mass Calculation model).

Fig. 3 is the establishment and the modification process of parametrization octane value computation model.

Fig. 4 is that parametrization gasoline octane rating Model Calculation parameter is used flow process (the main process of Model Calculation) figure.

Fig. 5 is flow process (model parameter study, the main process of the adjusting) figure of parametrization gasoline octane rating Model Calculation parameter adjustment.

Fig. 6 is cost optimization method flow (cost optimization algorithm) figure.

Fig. 7 is quantity optimization method idiographic flow (quantity optimization algorithm) figure.

Fig. 8 is profit optimization method idiographic flow (profit optimization algorithm) figure.

Embodiment

Below in conjunction with drawings and Examples the present invention is described in further detail.

Overall system structure of the present invention as shown in Figure 1, comprise that the input of blending information, octane value calculate, concoct formulation optimization calculating, the visual output of scheme, parameter on-line study, the information from the user is accepted in the input of wherein blending information, the index of component oil, the index of processed oil and to the evaluation and the selection of various blending schemes, and deposit information in corresponding database; It is to go out basis blending prescription and quality index according to the octane value in the component oil index, construction parametrization octane value model and associated information calculation that described octane value calculates, and deposit data the octane value database of prediction in, carry out the basis of formulation optimization as next step; It is resource constraint and market demand in the each economic target, parametrization gasoline octane value model, factory according to component oil, processed oil that described blending formulation optimization calculates, with the maximum economic benefit is the quantity (blending prescription) that target provides each component oil of participating in blending, and deposits related data in blending scheduling scheme database; The visual output of described scheme is that the blending scheduling scheme is outputed on the computer screen with visual phraseologies such as chart, process flow sheet, the layout drawings that installs pipes, for dispatcher's reference; Described parameter on-line study is the error by error, practical solution and the theoretical scheme that compares actual octane value and calculated octane number, with the method for machine learning model parameter is carried out intellectuality adjustment, and the adjustment result deposits the model parameter database in.

In the present invention program, input of blending information and the visual output of scheme are the input and output of system, and octane value calculates, formulation optimization calculates, the parameter on-line study is the key of technology; A good gasoline concoction prescription will satisfy two conditions: gasoline octane rating and optimum economic benefit.For calculated octane number, use the gasoline octane value model, and the intellectuality adjustment that model carries out along with concrete gasoline producing process, the crude oil place of production etc. again.Therefore, parametrization gasoline octane value model, the intelligent inflation method of gasoline octane rating model parameter, formulation optimization method content is a key of the present invention, illustrates especially below.

1) parametrization gasoline octane value model

Parametrization gasoline octane value model is the basis of gasoline octane rating intelligence blending, and it comes from down the octane value model of telling new will in person:

$R_m=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{X}_i\frac{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{X}_j{R}_{\mathrm{ij}}{Q}_{\mathrm{ij}}}{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{X}_j{Q}_{\mathrm{ij}}}---\left(1\right)$

Wherein:

R _mThe blended gasoline octane value;

X _jThe massfraction of component j;

R _IjComponent i (is generally (R to the octane value of component j _i+ R _j)/2);

Q _Ij(it is according to X to component i to the operational factors of component j _iAnd X _jSet).

Old model has only been considered the influence of the difference of component oil type (component oil i and component oil j) to blending octane value to Q, and consider since the variation of component oil ratio to the influence of blending octane value, so there is certain restriction in suitability.The present invention has made the function that definition again: Q is not only the component oil type to Q, also is the dispersing function of relative proportion between the component oil, i.e. Q simultaneously _Ij=Q (X _Ij, X _Ji), to become a kind of be the adjustable parameter model of variable with the component oil ratio to model like this.

2) the intelligent inflation method of gasoline octane rating model parameter for accurately describing the influence to octane value of blended gasoline octane value and component oil ratio variation relation and component oil type, has adopted a kind of intelligentized on-line study method to come the adjustment model parameter.

As calculated octane number R _cWith actual octane value R _fGap Δ R=R _f-R _c, in the time of not in allowed band (the permissible error limit that 0＜Δ R＜user sets), parameter Q learnt and adjust according to error.

The concrete grammar of adjusting is:

Q _ij(k+1)＝Q _ij(k)+β*ΔQ _ij(k) (2)

Here,

$\mathrm{\Δ}{Q}_{\mathrm{ij}}=\mathrm{\ΔR}\frac{\∂R}{\∂Q_{\mathrm{ij}}}$

Formula 2 is the formulas that iterate.Wherein, k represents to learn number of times; 0≤β≤1 is the study step-length; Δ R=R _f-R _cBe laboratory octane number R _fWith theoretical octane value R _cPoor; Δ Q _IjBe Q _IjIncrement, Be that R is to Q _IjPartial differential (can calculate) according to the formula 1 that provides above, as Δ Q _IjNear zero, Q _IjStudy finishes when (k+1) tending towards stability.

Model parameter Q _IjInitial value the two kinds of different situations that exist are set, for existing component oil type, Q _IjInitial value can be provided with by rule of thumb data; To the new component oil type that added afterwards, initial value is provided with by 1, promptly presses linear model as the initial calculation model.The main process of calculating is seen Fig. 4, Fig. 5.

The idiographic flow that described parametrization gasoline octane rating Model Calculation parameter is used is referring to Fig. 4, calculate the relative proportion between each component oil, judge and check the model parameter Q that whether exists between the component oil in twos through study, there are not then 2 study by formula of model parameter Q, carry out blending octane value again and calculate through study; As have model parameter Q, then directly calculating parameter Q is set by initial value.

The idiographic flow of described parametrization gasoline octane rating Model Calculation parameter adjustment as shown in Figure 5, to newly fill a prescription as current formulation, each component oil type and ratio in current formulation are provided with coefficient of mutual influence, the number of times of coefficient of mutual influence by study sorted from less to more, calculate the octane value of current formulation, allow the prescription octane value difference actual octane value that equals to fill a prescription deduct the formula calculation octane value, judge whether prescription octane value difference meets the demands, do not meet the demands, select one group of coefficient of mutual influence as adjusting object, and according to prescription octane value difference it is adjusted, return back to the octane value that calculates current formulation; Meet the demands, judge again whether all prescriptions all meet the demands, and all do not meet the demands as all prescriptions, return back to the octane value that calculates current formulation after selecting next assembly side as current formulation again; All meet the demands as all prescriptions, then parameter learning, adjustment algorithm finish.

3) gasoline concoction formulation optimization method

Parametrization gasoline octane value model and inflation method thereof have guaranteed that prescription can satisfy the gasoline octane rating index request, and formulation optimization will guarantee that then the petroleum products blending process satisfies the requirement of economic target.Because the influence of oil cost factor, refining process factor and other correlative factors, the cost that cause different times, obtains dissimilar component oils by different refining process all has very big difference, and this makes that same prescription is that the economic benefit that enterprise brought also is different in different periods; Both made in the same period, and under same quality index requires, when two or more component oils is concocted, multiple blending scheme had been arranged usually, and the economic benefit that every kind of blending scheme is brought also there is very big difference.Therefore need to make that by optimization of C under any condition, the component oil of different mass, different costs is able to sufficient utilization, creates more value.

In formulation optimization process in the past, the main method that adopts is that the octane value model is formed the system of equations of calculating economic benefit after with various constraint condition simultaneous, uses then that method is optimized calculating in the operational research.Because in whole process, the octane value model participates in formulation optimization together with the economic benefit condition and calculates, this makes the octane value model very close with the economic benefit conditional relationship, can be called a kind of " tight coupling " relation.The method of utilizing " tight coupling " to concern is carried out formulation optimization, and there are the following problems:

Parametrization octane value model is a kind of series model that is formed by component oil type and their proportionlity in fact, and non-traditional single model.Component oil coefficient of mutual influence Qij is the dispersing function of relative proportion between component oil type and the component oil.When carrying out formulation optimization, need preestablish Qij, be optimized calculating again by " tight coupling " mode.And the relative proportion of each component oil of filling a prescription after the process computation optimization does not match with the Qij that sets most probably, and at this moment octane value can depart from preset value, and optimum result can't satisfy needs of production.

For addressing this problem, the present invention proposes the way of thinking of carrying out the petroleum products blending formulation optimization by " loose coupling ": provide one group of blending prescription by experience earlier, prescription is optimized processing, verify according to the octane value computation model whether this prescription meets index request again, as the then end that meets the requirements is calculated, otherwise calculation result is returned to the formulation optimization process, carry out the reference or the basis of next step optimization as the computation optimization model.Relation between computation optimization model and the Mass Calculation model is shown in Fig. 2,3.

Total system has study and calculates two states: learning state mainly is study and adjusts calculating parameter; Working order is then carried out octane value calculating, computation optimization, design blending scheme etc. according to the parameter that learns.

For ease of calculating, the present invention is divided into two kinds of high-octane rating and low octane ratings to component oil, high-octane number component oil is meant that octane value is higher than the component oil of target octane value (processed oil label), and in like manner, the low octane rating component oil is meant that octane value is lower than the component oil of target octane value (processed oil label).

Actual demand according to oil refining is produced the present invention proposes cost optimization method, quantity optimization method and three kinds of formulation optimization methods of profit optimization method (seeing accompanying drawing 6,7,8).Wherein, the cost optimization method is target with the least cost, and the quantity optimization method is target with the maximum output, and the profit optimization method is target with the maximum profit.The target of these three kinds of formulation optimization methods is different, and the scheduling decision person can select a best blending scheme according to actual needs voluntarily.

The flow process of described cost optimization method as shown in Figure 6, described cost optimization method idiographic flow is: earlier component oil is sorted from small to large by octane value, find the dividing point of high-octane number component oil and low octane rating component oil; The component oil that the user is chosen just is divided into two groups according to octane value again, make up respectively in twos, and octane number requirement according to target, the cost that calculates every kind of combination reaches the wherein ratio of various component oils, and with minimum one group of cost as basic optimum combination, never take out a kind of conduct in the component oil in optimizing the basis combination and newly add component oil, from optimize the basis combination, select each a kind of component oil of height as displacement component oil, to newly add component oil replaces by mode such as equal proportion such as cost such as grade with displacement component oil, judge then whether the result after the displacement can carry out cost optimization, as carrying out cost optimization, then increase the new ratio that adds component oil, judge whether existing component oil reaches the ratio restriction in three kinds of component oils, reach the ratio restriction then with the component oil in the current component oil note do optimization basis combination, judge whether again also to exist and do not add the component oil of optimizing the basis combination, be to take out a kind of conduct in the component oil that then is back to never in optimizing the basis combination newly to add component oil, select each a kind of component oil of height as displacement component oil in the combination of basis from optimizing, otherwise the cost optimization algorithm finish; Do not reach the ratio restriction, be back to and replace by mode such as equal proportion such as cost such as grade with displacement component oil newly adding component oil; As carrying out cost optimization, then adjust the ratio of displacement component oil, make basic quality index of compose of panel just reach target call, calculate this cost, whether the cost of doing this then is less than the judgement of the optimization cost of last time, when the optimization cost that is not less than last time up to increasing the new ratio that adds component oil, the ratio of current each component oil of record is as optimized proportion, again to increasing the new ratio that adds component oil when less than the optimization cost of last time.

As described in Figure 7, described quantity optimization method idiographic flow is: component oil is sorted from small to large by octane value, find the dividing point of high-octane number component oil and low octane rating component oil, note is according to the tank farm stock of each component oil, blending octane value when calculating all components oil is concocted together, the blending result is higher than requirement and then turns to the component of high group is handled, and meets the demands to the quantity maximization up to the octane value that calculates, and withdraws from the computation optimization program; The blending result is lower than requirement and then turns to the component of low group is handled, and meets the demands up to the octane value that calculates, and to the quantity maximization, withdraws from the computation optimization program; The blending result meets the demands and then directly carries out the quantity maximization, withdraws from the computation optimization program.

As described in Figure 8, described profit optimization method idiographic flow is: with the target octane value is boundary, choose the user to such an extent that component oil just is divided into two groups according to octane value, again the component oil of height in two groups carried out combinations of pairs in twos, according to target octane value calculates ratio separately, and with tank farm stock separately as usage quantity but (residual content), but then with usage quantity (residual content) separately, calculate every kind of obtainable profit of combination, choose a kind of combination back of profit maximum and judge whether maximum profit is less than or equal to 0, maximum profit is not less than usage quantity and the residual content calculated after this group component oil of choosing is concocted in this ratio at 0 o'clock, the usage quantity of combination is newly added in total usage quantity of prescription, component oil mass to the quality index of adjusting prescription meets the requirements of scope, return back to then with tank farm stock separately as usage quantity but (residual content), but, finished profit as maximum profit less than 0 o'clock and optimize algorithm then with usage quantity (residual content) separately.

Claims (10)

1. the intelligent blending method of a gasoline octane rating, it is characterized in that: comprise that the input of blending information, octane value calculate, concoct formulation optimization calculating, the visual output of scheme, parameter on-line study, the information from the user is accepted in the input of wherein blending information, and deposits information in corresponding database; Described octane value calculate be according to component oil octane value, build parametrization octane value model and associated information calculation goes out basis blending prescription and quality index, and deposit data the octane value database of prediction in, carry out the basis of formulation optimization as next step; It is resource constraint and market demand in the each economic target, parametrization gasoline octane value model, factory according to component oil, processed oil that described blending formulation optimization calculates, with the maximum economic benefit is the quantity that target provides each component oil of participating in blending, and deposits related data in blending scheduling scheme database; The visual output of described scheme is exported with visual phraseology; Described parameter on-line study is by the error of error, practical solution and the theoretical scheme of actual octane value and calculated octane number relatively, with machine learning method model parameter is carried out intellectuality adjustment, and will adjust the result and deposit the model parameter database in.

2. according to the intelligent blending method of the described gasoline octane rating of claim 1, it is characterized in that: described parametrization gasoline octane value model is the basis of gasoline octane rating intelligence blending, and it comes from following octane value model, and is variable with the component oil ratio:

$R_m=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{X}_i\frac{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{X}_j{R}_{\mathrm{ij}}{Q}_{\mathrm{ij}}}{\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{X}_j{Q}_{\mathrm{ij}}}----\left(1\right)$

Wherein:

R _mThe blended gasoline octane value;

X _jThe massfraction of component j;

R _IjComponent i is to the octane value of component j;

Q is not only the function of component oil type, also is the dispersing function of relative proportion between the component oil, i.e. Q simultaneously _Ij=Q (X _Ij, X _Ji), making described is the adjustable parameter model of variable with the component oil ratio.

3. according to the intelligent blending method of the described gasoline octane rating of claim 2, it is characterized in that: described intelligentized on-line study method is used for accurately describing the influence to octane value of blended gasoline octane value and component oil ratio variation relation and component oil type, and parametrization gasoline octane rating Model Calculation method is:

As calculated octane number R _cWith actual octane value R _fGap Δ R=R _f-R _c, in the time of not in allowed band, parameter Q learnt and adjust according to error; The concrete grammar of adjusting is:

Q _ij(k+1)＝Q _ij(k)+β*ΔQ _ij(k) (2)

Here,

$\mathrm{\Δ}{Q}_{\mathrm{ij}}=\mathrm{\ΔR}\frac{\∂R}{\∂Q_{\mathrm{ij}}}$

Formula 2 is the formulas that iterate; Wherein k represents to learn number of times; 0≤β≤1 is the study step-length; Δ R=R _f-R _cBe laboratory octane number R _fWith theoretical octane value R _cPoor; Δ Q _IjBe Q _IjIncrement,

Be that R is to Q _IjPartial differential, as Δ Q _IjNear zero, Q _IjStudy finishes when (k+1) tending towards stability.

4. according to the intelligent blending method of the described gasoline octane rating of claim 3, it is characterized in that: described allowed band is the permissible error limit that 0＜Δ R＜user sets.

5. according to the intelligent blending method of the described gasoline octane rating of claim 2, it is characterized in that: the idiographic flow that described parametrization gasoline octane rating Model Calculation parameter is used is: calculate the relative proportion between each component oil, judge and check the model parameter Q that whether exists between the component oil in twos through study, there are not then 2 study by formula of model parameter Q, carry out blending octane value again and calculate through study; As have model parameter Q, then directly calculating parameter Q is set by initial value.

6. according to the intelligent blending method of the described gasoline octane rating of claim 2, it is characterized in that: the idiographic flow of described parametrization gasoline octane rating Model Calculation parameter adjustment is: will newly fill a prescription as current formulation, each component oil type and ratio in current formulation are provided with coefficient of mutual influence, the number of times of coefficient of mutual influence by study sorted from less to more, calculate the octane value of current formulation, allow the prescription octane value difference actual octane value that equals to fill a prescription deduct the formula calculation octane value, judge whether prescription octane value difference meets the demands, do not meet the demands, select one group of coefficient of mutual influence as adjusting object, and according to prescription octane value difference it is adjusted, return back to the octane value that calculates current formulation; Meet the demands, judge again whether all prescriptions all meet the demands, and all do not meet the demands as all prescriptions, return back to the octane value that calculates current formulation after selecting next assembly side as current formulation again; All meet the demands as all prescriptions, then parameter learning, adjustment algorithm finish.

7. according to the intelligent blending method of the described gasoline octane rating of claim 1, it is characterized in that: described gasoline concoction formulation optimization method is undertaken by " loose coupling " petroleum products blending formulation optimization thought: provide one group of blending prescription by experience earlier, prescription is optimized processing, verify according to octane value calculating parameter gasoline octane value model whether this prescription meets index request again, as the then end that meets the requirements is calculated, otherwise calculation result is returned to described gasoline concoction formulation optimization process, as further reference or the basis of optimizing of computation optimization model; Specifically can adopt cost optimization method, quantity optimization method and profit optimization method.

8. according to the intelligent blending method of the described gasoline octane rating of claim 7, it is characterized in that: described cost optimization method idiographic flow is: earlier component oil is sorted from small to large by octane value, find the dividing point of high-octane number component oil and low octane rating component oil; The component oil that the user is chosen just is divided into two groups according to octane value again, make up respectively in twos, and octane number requirement according to target, the cost that calculates every kind of combination reaches the wherein ratio of various component oils, and with minimum one group of cost as basic optimum combination, never take out a kind of conduct in the component oil in optimizing the basis combination and newly add component oil, from optimize the basis combination, select each a kind of component oil of height as displacement component oil, to newly add component oil replaces by mode such as equal proportion such as cost such as grade with displacement component oil, judge then whether the result after the displacement can carry out cost optimization, as carrying out cost optimization, then increase the new ratio that adds component oil, judge whether existing component oil reaches the ratio restriction in three kinds of component oils, reach the ratio restriction then with the component oil in the current component oil note do optimization basis combination, judge whether again also to exist and do not add the component oil of optimizing the basis combination, be to take out a kind of conduct in the component oil that then is back to never in optimizing the basis combination newly to add component oil, select each a kind of component oil of height as displacement component oil in the combination of basis from optimizing, otherwise the cost optimization algorithm finish; Do not reach the ratio restriction, be back to and replace by mode such as equal proportion such as cost such as grade with displacement component oil newly adding component oil; As carrying out cost optimization, then adjust the ratio of displacement component oil, make basic quality index of compose of panel just reach target call, calculate this cost, whether the cost of doing this then is less than the judgement of the optimization cost of last time, when the optimization cost that is not less than last time up to increasing the new ratio that adds component oil, the ratio of current each component oil of record is as optimized proportion, again to increasing the new ratio that adds component oil when less than the optimization cost of last time.

9. according to the intelligent blending method of the described gasoline octane rating of claim 7, it is characterized in that: described quantity optimization method idiographic flow is: component oil is sorted from small to large by octane value, find the dividing point of high-octane number component oil and low octane rating component oil, note is according to the tank farm stock of each component oil, blending octane value when calculating all components oil is concocted together, the blending result is higher than requirement and then turns to the component of high group is handled, meet the demands to the quantity maximization up to the octane value that calculates, withdraw from the computation optimization program; The blending result is lower than requirement and then turns to the component of low group is handled, and meets the demands up to the octane value that calculates, and to the quantity maximization, withdraws from the computation optimization program; The blending result meets the demands and then directly carries out the quantity maximization, withdraws from the computation optimization program.

10. according to the intelligent blending method of the described gasoline octane rating of claim 7, it is characterized in that: described profit optimization method is boundary with the target octane value, choose the user to such an extent that component oil just is divided into two groups according to octane value, again the component oil of height in two groups carried out combinations of pairs in twos, according to target octane value calculates ratio separately, but and with tank farm stock separately as usage quantity, but then with usage quantity separately, calculate every kind of obtainable profit of combination, choose a kind of combination back of profit maximum and judge whether maximum profit is less than or equal to 0, maximum profit is not less than usage quantity and the residual content calculated after this group component oil of choosing is concocted in this ratio at 0 o'clock, the usage quantity of combination is newly added in total usage quantity of prescription, component oil mass to the quality index of adjusting prescription meets the requirements of scope, but return back to then with tank farm stock separately as usage quantity, but, finished profit as maximum profit less than 0 o'clock and optimize algorithm then with usage quantity separately.

Images