CN105727777A - Heavy bunker fuel oil optimization blending method - Google Patents

Abstract

The invention relates to a heavy bunker fuel oil optimization blending method. The method comprises 1, detecting and analyzing all indexes of raw oil for fuel oil blending, 2, building an objective function, 3, building a sample set, 4, building a formula model, 5, optimizing the formula model and 6, outputting the blending formula obtained by the step 5 to the next process control system and executing a process. The heavy bunker fuel oil optimization blending method can screen the optimal formula in compositions of multiple types of blending raw oil products, satisfies national quality standard requirements, has a small raw oil blending cost and less detection workload and greatly improves screening efficiency.

Description

A kind of heavy bunker fuel oil beending optimization method

Technical field

The present invention relates to a kind of heavy bunker fuel oil blending method.The cycle obtaining blending formula can be shortened by the method, improve the utilization ratio of Blend Stocks inferior, reduce the blending cost of bunker fuel oil.

Background technology

Along with the high speed development of world's marine trade, heavy bunker fuel oil demand constantly increases.But marine fuel oil yield that refinery is directly produced is relatively low, far from meeting the market demand.Accordingly, it would be desirable to exploitation blending technology, part industrial waste oil is mixed into bunker fuel oil, to ensure shipping demand.

Due to the progress of refinery technology, the kind of the raw oil being currently available for the blending of heavy bunker fuel oil gets more and more, thus causing that potential blending scheme increases in a large number.Look for least cost blending scheme by conventional experience or exhaustive method and need to expend substantial amounts of experiment detection work, extremely inefficient, it is difficult to adapt to the Rapid Variable Design of fuel oil market, affect enterprise getting profit level.

The current research about petroleum products blending technology focuses mostly in product oil field, and owing to the raw materials used oil nature of Product Oil Blending is more stable, nonlinear indicator forecast model is also comparatively perfect.Therefore, Product Oil Blending is paid attention to solve optimum blending proportioning in the feasible zone that known index prediction model limits.Such as document CN103065204A discloses the gasoline concoction optimization of a kind of VERNA-GA, and the method is mainly used in when the index prediction models such as octane number have built up, and uses genetic algorithm for solving blending formula, reaches benefit；The indexs such as octane number used by CN103745115A and CN101694571A are also comparatively clear and definite.

But the correlational study achievement utilizing multi items raw oil blending bunker fuel oil is not yet ripe.Because heavy oil peculiar to vessel concocts raw materials used oil is mainly low grade oils, price fluctuation is big, the oil product index of the same race of different batches also has obvious difference, and the feedstock property difference of different cultivars is bigger, and the forecast model of the nonlinear indicator such as viscosity, pour point is also indefinite.That is, the feasible zone of blending proportioning is indefinite.For the beending optimization of heavy oil peculiar to vessel, not only want calculation optimization proportioning, it is often more important that clearly optimize feasible zone, namely also constantly to improve index prediction model by great many of experiments detection data in optimization process.Therefore, what the present invention focused on is shorten improve index prediction model and obtain the cycle of optimization formula, to meet the ageing of fuel oil market price.

Summary of the invention

In order to solve above-mentioned technical problem, the invention provides a kind of fuel oil beending optimization method, it is possible to reach to reduce experiment detection analytical work amount, reduce fuel oil blending cost and improve the purpose of poor energy utilization ratio.

To achieve these goals, the present invention adopts the following technical scheme that

A kind of heavy bunker fuel oil beending optimization method, comprises the steps:

Step (1): the indices of the raw oil being in harmonious proportion for fuel oil is analyzed in detection；

Step (2): the foundation of object function；

Step (3): the foundation of sample set；

Step (4): the foundation of formulation model；

Step (5): the optimization of formulation model；

Step (6): blending formula output to downstream step (5) obtained controls system and performs.

Adopting blending method of the present invention, it is possible to effectively utilize poor energy blended fuel oil, and the blending cycle is short, blending cost is high.

As the preferred embodiment of the present invention, in the above-mentioned methods,

In step (1), described index refers to the quality index that the fuel oil of national Specification must is fulfilled for.

In step (2), described object function is fuel oil blending cost minimum function, namely tried to achieve in Board Lot blended fuel oil by function the ratio shared by each raw oil and respective procurement price product and minimum, namely:

$Y=\min \left(\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{P}_i{x}_i\right)---\left(1\right)$

Wherein:

N is the kind of raw oil, n=1,2,3, i；

P_iIt it is the market price of i-th kind of raw oil；

x_iIt is the blending ratio shared by i-th kind of raw oil,

Y is blending cost.

In step (3), also specifically include following steps:

1) stochastic generation blending formula sample；

2) sample is carried out economy screening；

Adopting weighted average method to obtain the mediation cost of formula sample, if blending cost is lower than the market price, then has profit and be likely to, economy is qualified；

If formula sample cost is higher than the market price, then abandons this sample, stochastic generation next one formula sample, carry out cost calculation.

3) to meeting step 2) sample carry out the screening of linear index；

Utilize weighted average method to obtain every linear index of formula sample, as sulfur content, water content and various tenors etc. limit index, if index meets national standard, then enter next step screening；

Index such as formula sample does not meet national standard, then abandon this sample, regenerates formula sample, until linear index all reaches requirement；

4) to step 3) sample carry out the screening of similarity

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}|x_i-x_j|>\mathrm{Similar},1\&le;j<i,i\&le;N---\left(4\right)$

Similarity is set and controls parameter；After generating i-th sample, start the similarity checking each sample successively with newly-generated i sample from first sample (j=1), if any one similarity controls parameter less than similarity, then abandon this sample, and regenerate.

By meeting above-mentioned linear index, the formula sample of cost and similarity of being in harmonious proportion be stored in sample plan collection and continue to generate next sample and repeat above-mentioned screening, until the total sample number that sample plan is concentrated reaches 10* (n-1)²Till, n is the species number of raw oil.

In step (4), comprise the steps:

1) detection sample plan concentrates the viscosity of formula sample oil product, density, pour point and flash-point, and calculates the carbon aromaticity index of each sample；

2) sample in scheme sample set is trained, thus obtaining the training tool predictive ability to viscosity, density, pour point, flash-point and carbon aromaticity index index；

Particularly as follows: sample and viscosity, density, pour point and flash-point in Utilization plan sample set limit Indexs measure result, the nonlinear indicator forecast models such as the viscosity in restrictive condition are trained, obtain each coefficient limiting index prediction model in constraints, thus improving restrictive condition further；Described training tool is preferably the instrument such as artificial neural network or support vector machine；

3) preferred blending scheme is solved；Described method for solving is preferably genetic algorithm or flock of birds algorithm.

In step (5), the preferred blending scheme obtained in detecting step (4) carries out viscosity, density, pour point, flash-point and carbon aromaticity index index；

If testing result be unsatisfactory for national standard or testing result meets national standard but optimizing algebraically is not up to setting value (as manually set optional 100 generations), then optimizing sample is returned to sample set, utilize new sample set to repeat step (4) and revise formulation model, until gained blending formula meets optimizes end condition.

Adopt heavy bunker fuel oil beending optimization method of the present invention, can comform comparatively quickly multi items Blend Stocks oil proportioning combination in screening optimization formula, under meeting the premise that state quality standard requires, make fuel oil blending cost minimization, detection workload is minimum, and screening efficiency is greatly improved.

Accompanying drawing explanation

Fig. 1 is bunker fuel oil beending optimization method flow diagram.

Detailed description of the invention

Following example are used for illustrating the present invention, but are not limited to the scope of the present invention.

The beending optimization method of 1 one kinds of heavy bunker fuel oils of embodiment

Fuel oil to concoct No. 180#:

Step one, detection analyze the indices of the raw oil being in harmonious proportion for 180# fuel oil

Raw oil is selected specifically to: shale oil, oil waterborne, coal diesel oil, slurry oil, Liaohe River Colophonium, totally 5 kinds of raw oil.Each raw oil is refinery's waste oil, and its performance indications are as shown in table 1.

Table 1 is for the performance indications of each raw oil concocted and price

Step 2: the foundation of object function

Object function is that fuel oil blending cost is minimum, namely in Board Lot blended fuel oil the product of ratio shared by each raw oil and respective procurement price and be minimum, namely:

$Y=\min \left(\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{P}_i{x}_i\right)$

Wherein: x_iIt is the blending ratio shared by i-th kind of raw oil,P_iIt it is the market price of i-th kind of raw oil.

Step 3: the foundation of sample set

1) stochastic generation blending formula sample 1:

Shale oil 14%, oily 0%, coal diesel oil 13%, slurry oil 34%, Liaohe River Colophonium 39% waterborne.

2) sample is carried out economy screening

With market price Price_Profit=4600 is standard, and the cost of formula sample 1 is: $\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{P}_i{x}_i=5000\&times;0.14+3300\&times;0.13+4100\&times;0.34+4500\&times;0.39=4278,$ Less than the market price, having profit and be likely to, economy is qualified.

If the cost of formula sample 1 is higher than the market price, by stochastic generation formula 2, carry out cost calculation.

3) to meeting step 2) sample carry out the screening of linear index:

About sulfur content less than 3.5% in GB GB/T17411-2012, as standard, calculate the sulfur content in formula sample 1.

, less than national regulations sulfur content, qualified.

Other hydrogen sulfide content, acid number, total precipitate, carbon residue, moisture, ash, content of vanadium, sodium content, aluminum+silicone content, calcium content, Zn content, phosphorus content totally 12 indexs are similar to sulfur content index screening mode, comparing with GB, qualified enters next step screening；

If defective, then abandon this sample, regenerate sample, until linear index all reaches requirement；

4) to step 3) sample that obtains carries out the screening of similarity:

Sample Similarity is set and controls parameter Similar=30, Similar=Similar-δ * 5, $\mathrm{\&delta;}=\left\{\begin{array}{cc}0,& \mathrm{gn}<2000\\ 1,& \mathrm{gn}=2000\end{array}\right.,$ Wherein gn represents the number of times randomly generating i-th sample, gn=0 time initial, and when often producing a defective sample, gn adds 1, when namely gn=2000 or gn < 2000 successfully generate qualified i-th sample, again makes gn=0.

If any one similarity controls parameter less than similarity, then abandon this sample, regenerate sample, until similarity all reaches requirement；

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}100\&times;|x_i-x_j|>\mathrm{Similar},1\&le;j<i,i\&le;n;$

Using meeting above-mentioned linear index, the formula of cost and similarity of being in harmonious proportion be stored in sample plan collection as sample and continue to generate next sample and repeat above-mentioned screening, until the total sample number that sample plan is concentrated reaches 10* (n-1)²Till.

Step 4, formulation model foundation, i.e. the foundation of nonlinear indicator Forecasting Methodology:

1) the scheme blending sample concentrated by sample plan, the viscosity of oil product, density, pour point and the flash-point that detection is concocted out, and specify to calculate the carbon aromaticity index of each blending sample by GB GB/T17411-2012；

2) utilize the instrument such as artificial neural network or support vector machine that the sample in blending scheme sample set is trained, thus obtaining artificial neural network or the support vector machine instrument predictive ability to viscosity, density, pour point, flash-point and carbon aromaticity index index；

3) genetic algorithm, flock of birds algorithm etc. is selected to solve preferred blending scheme.

Step 5, formulation model optimization:

Preferred blending scheme step 4 obtained carries out the detection of viscosity, density, pour point, flash-point and carbon aromaticity index index；If testing result be unsatisfactory for national standard or testing result meets national standard but optimizing algebraically (manually sets not up to setting value, optional 100 generations), then optimizing sample is added into sample set, new sample set is utilized to repeat step 4 correction formulation model, until gained blending formula meets optimizes end condition.

When, after all index conformance with standard of blending scheme optimized, formula is concocted in output:

Shale oil: oil waterborne: coal diesel oil: slurry oil: Liaohe River Colophonium

=0.15:0.10:0.30:0.15:0.30.

The fuel oil quality finally given meets national Specification, and has the formula of Optimum Economic benefit.

Step 6, blending formula step 5 obtained output to downstream controls system and performs.

By contrast, use existing blending method allotment 180# fuel oil, 90～120 days acquisition cycles and adopt blending method of the present invention obtain 180# fuel oil, obtain 30～45 days cycles.

Although, above the present invention is described in detail with a general description of the specific embodiments, but on basis of the present invention, it is possible to it is made some modifications or improvements, and this will be apparent to those skilled in the art.Therefore, these modifications or improvements without departing from theon the basis of the spirit of the present invention, belong to the scope of protection of present invention.

Claims (8)

1. a heavy bunker fuel oil beending optimization method, it is characterised in that comprise the steps:

Step (1): the indices of the raw oil being in harmonious proportion for fuel oil is analyzed in detection；

Step (2): the foundation of object function；

Step (3): the foundation of sample set；

Step (4): the foundation of formulation model；

Step (5): the optimization of formulation model；

Step (6): blending formula output to downstream step (5) obtained controls system and performs.

2. heavy bunker fuel oil beending optimization method according to claim 1, it is characterised in that in step (2), described object function is fuel oil blending cost minimum function.

3. heavy bunker fuel oil beending optimization method according to claim 1, it is characterised in that described step (3) specifically includes following steps:

1) stochastic generation blending formula sample；

2) sample is carried out economy screening；

3) to meeting step 2) sample carry out the screening of linear index；

4) to meeting step 3) sample carry out the screening of similarity；

By meeting above-mentioned linear index, the formula sample of cost and similarity of being in harmonious proportion be stored in sample plan collection and continue to generate next sample and repeat above-mentioned screening, until the total sample number that sample plan is concentrated reaches 10* (n-1)²Till.

4. heavy bunker fuel oil beending optimization method according to claim 3, it is characterised in that described step 4) similarity screening particularly as follows:

$\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}|x_i-x_j|>\mathrm{Similar},$ 1≤j < i, i≤N；

Similarity is set and controls parameter；After generating i-th sample, start the similarity checking each sample successively with newly-generated i sample from first sample (j=1), if any one similarity controls parameter less than similarity, then abandon this sample, and regenerate.

5. heavy bunker fuel oil beending optimization method according to claim 1, it is characterised in that described step (4) comprises the steps:

1) detection sample plan concentrates the viscosity of formula sample oil product, density, pour point and flash-point, and calculates the carbon aromaticity index of each sample；

2) sample in scheme sample set is trained, thus obtaining the training tool predictive ability to viscosity, density, pour point, flash-point and carbon aromaticity index index；

3) preferred blending scheme is solved.

6. heavy bunker fuel oil beending optimization method according to claim 5, it is characterised in that described training tool is artificial neural network or support vector machine.

7. heavy bunker fuel oil beending optimization method according to claim 5, it is characterised in that described method for solving is genetic algorithm or flock of birds algorithm.

8. heavy bunker fuel oil beending optimization method according to claim 1, it is characterized in that, in described step (5), the preferred blending scheme obtained in detecting step (4) carries out viscosity, density, pour point, flash-point and carbon aromaticity index index；

If testing result be unsatisfactory for national standard or testing result meets national standard but optimizing algebraically is not up to setting value, then sample is returned to sample set, utilize new sample set to repeat step (4) and revise formulation model, until gained blending formula meets optimizes end condition.