CN108415246B - Crude oil nonlinear optimization blending method based on expanded initialization range - Google Patents

Abstract

The invention discloses a crude oil nonlinear optimization blending method based on expanded initialization range, which comprises the following steps: expanding the formula range of m components participating in blending crude oil to generate an expanded formula range, generating 3m groups of initial mass part ratios in the expanded formula upper limit range, the expanded formula lower limit range, the reduced formula range and the expanded formula range, further performing optimization solution to obtain an optimized formula of the component crude oil, calculating the feasibility of each optimized formula solution, forming a set of optimized formula solutions from feasible optimized formula solutions of the component crude oil, and selecting an optimal group of solutions as a final formula of the component crude oil according to the size of a target function value in a crude oil nonlinear blending optimization algorithm. The crude oil nonlinear optimization blending method based on partition initialization solves the problem that the optimization property target is easy to fall into a local extreme value or cause no solution in optimization when a nonlinear blending rule is adopted in the crude oil online blending process.

Description

Crude oil nonlinear optimization blending method based on expanded initialization range

Technical Field

The invention relates to the field of crude oil online blending of oil refining enterprises, in particular to a situation that the nonlinear blending of crude oil falls into a local extreme value or is optimized without solution.

Background

In a constraint optimization problem in real-world application, an optimal solution is often located on or near a constraint boundary, or a feasible region occupies a small proportion of the whole search space (composed of the feasible region and the infeasible region). In this case, the position of the infeasible solution near the optimal solution is likely to have a reference value more than the position of the feasible solution located inside the feasible domain, and how to sufficiently utilize the infeasible solution to find the global optimal solution of the constraint optimization problem is very important.

In the petroleum and petrochemical industry, the crude oil on-line blending optimization technology is successfully applied in recent years. The technology mainly aims at controlling indexes such as sulfur content, acid value, naphtha yield and the like of crude oil at present, reduces the corrosion degree of a CDU device, enables the device to run stably, and simultaneously guarantees partial processing requirements of a secondary production device.

However, when the optimization properties include non-linear blending rules such as octane number, the situation of local extrema trapping or no solution for optimization often occurs, mainly because the optimal solution is often located on or near the constraint boundary, or the feasible region occupies a small proportion of the whole search space (composed of the feasible region and the infeasible region). This can be caused when no feasible solution can be given: 1) the properties of the blended oil exceed the allowable range of a CDU device, so that the corrosion of the device is aggravated; 2) the properties of oil products such as produced slag oil and the like which need secondary processing exceed the standard and do not meet the requirements of secondary processing devices; 3) the produced product is unqualified and can not be sold out of the factory. Therefore, in order to meet the normal production of refineries, the solution of the problem is urgently needed.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides a crude oil nonlinear optimization blending method based on an expanded initialization range. The method mainly solves the problem that the nonlinear optimization target optimization has no solution or falls into a local extreme value in the crude oil online blending process.

The technical scheme is as follows: a crude oil nonlinear optimization blending method based on expanded initialization range comprises the following steps: expanding the formula range of m components participating in blending crude oil to generate an expanded formula range, generating 3m groups of initial mass part ratios in the expanded formula upper limit range, the expanded formula lower limit range, the reduced formula range and the expanded formula range, further performing optimization solution to obtain an optimized formula of the component crude oil, calculating the feasibility of each optimized formula solution, forming a set of optimized formula solutions from feasible optimized formula solutions of the component crude oil, and selecting an optimal group of solutions as a final formula of the component crude oil according to the size of a target function value in a crude oil nonlinear blending optimization algorithm. The problem that the crude oil is easy to fall into a local extreme value or is not solved in the optimization blending process is solved.

Preferably, in the blending optimization process of the crude oil, the blending optimization of each property is calculated, and a part or all of the properties adopt a nonlinear blending rule, wherein the octane number is calculated by adopting a secondary nonlinear regression model:

in the formula (1), f (x) is the octane number of the blended crude oil; m is the component number participating in crude oil blending; x is the number of_iIs the blending mass ratio of the component crude oil i; q_i,_jThe blending effect coefficient of blending component crude oil i and j, the size and the positive and negative of which reflect the nonlinear action between different crude oils, Q_i,_j≠Q_j,i，Q_i,i＝Q_j,j＝1；R_i,_jAs the arithmetic mean of the research octane numbers of the blended component crude oils i and j, R_i,_j＝(R_i+R_j)/2。

Preferably, the specific steps are as follows:

I. obtaining basic parameters of crude oil blending, including the number m of components participating in blending, the properties of the crude oil, and the formula range XR of each component_i(i ═ 1,2 … m), constraints such as an optimized property range;

II, establishing a crude oil nonlinear blending optimization equation;

III, limiting the formula upper limit x of each blending component_max,iAnd a lower limit x_min,iBoth are multiplied by (1 +%) and (1-%) to obtain the upper limit range [ x ] of the expanded formula_maxl,i,x_maxu,i]Expanding the lower limit range of the formula [ x_minl,i,x_minu,i]Reducing the formulation range [ x_minu,i,x_maxl,i]And rubbingExtended formulation range [ x_minl，i,x_maxu，i]；

In the range of expanding upper limit of formula [ x ]_maxl,i,x_maxu,i]Expanding the lower limit range of the formula [ x_minl,i,x_minu,i]Reducing the formulation range [ x_minu,i,x_maxl,i]And expand the formula range [ x_minl，i,x_maxu，i]Generating 3m groups of initial mass part ratios Xbe;

v, selecting a t-th group of new initial mass part ratios Xbe from the step IV_tInitially, t is 1;

VI, Xbe, with other blending basic parameters unchanged_tCalling a crude oil nonlinear blending optimization algorithm to solve an optimized formula XS of the component crude oil for initial mass part ratio_t；

VII, judging the optimized formula XS of the component crude oil solved in the step VI_tIf the solution is a feasible solution, entering the step VIII if the solution is a feasible solution, otherwise, directly entering the step IX;

VIII. mixing XS_tRecord to X_FSEntering the step IX;

if t is t +1, entering a step V when t < 3m, and otherwise, entering a step X;

x, ending the feasible solution process of the optimized formula of the crude oil;

XI after the process of obtaining a feasible solution is finished, compare X_FSThe optimal formula feasible solutions of all the crude oil components are selected as the final formula according to the value of the objective function in the crude oil nonlinear blending optimization algorithm.

Preferably, the upper limit x of the formula of each blending component is_max,iAnd a lower limit x_min,iBoth are multiplied by (1 +%) and (1-%) to obtain the upper limit range [ x ] of the expanded formula_maxl,i,x_maxu,i]And expanding the lower limit range of the formula [ x_minl,i,x_minu,i]The calculation process is as follows:

XR_i＝[x_min,i x_max,i]where i is 1,2 … m (2)

x_maxl，i＝x_max,i×(1-％) (3)

x_maxu，i＝x_max,i×(1+％) (4)

x_minl，i＝x_min,i×(1-％) (5)

x_minu，i＝x_min,i×(1+％) (6)

In formulae (2) to (6), XR_iFor blending the formula range of component i, x_max,iIs the maximum value of the formulation, x, of blending component i_min,iThe minimum value of the formula of the blending component i, m is the component number participating in crude oil blending, and the upper and lower limits of the formula of the blending component i are magnified and reduced and are between 0 and 10]In the meantime.

Preferably, in the range of the upper limit of the extended formula [ x ]_maxl，i,x_maxu，i]Expanding the lower limit range of the formula [ x_minl，i,x_minu，i]Narrow the range of the formula [ x_minu，i,x_maxl，i]And expanding the formulation range [ x ]_minl，i,x_maxu，i]The 3m set of initial recipes Xbe were generated and calculated as follows:

a) selecting the upper limit range [ x ] of the formulation of the ith component_maxl，i,x_maxu，i]Within this range, the mass fraction value X is randomly generated_bg,i-1With the remaining m-1 components in their extended formulation range [ x ]_minl，j,x_maxu，j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass ratio of Xbe_i，1；

b) Selecting the lower limit range [ x ] of the formulation of the ith component_minl，i,x_minu，i]Within this range, a value X is randomly generated_bg,i-2With the remaining m-1 components in their extended formulation range [ x ]_minl，j,x_maxu，j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass ratio of Xbe_i，2；

c) Selecting a reduced formula range [ x ] of the ith component_minu，i,x_maxl，i]Within this range, a value X is randomly generated_bg,i-3With the remaining m-1 components in their extended formulation range [ x ]_minl,j,x_maxu,j]All the other productions in the interiorGenerating 1 random number, and counting m-1 random number values X_bg,jUsing the m random numbers as a group to optimize the initial mass ratio of Xbe_i，3；

d) Repeating the steps a, b and c on all the components to generate 3m groups of initial mass part ratios Xbe.

Preferably, the objective function of the crude oil nonlinear blending optimization algorithm is as follows:

in the formula (7), Φ (X) is an objective function; h represents the number of properties of the crude oil participating in the optimization, f (X)_lBlending rules adopted for the first property of the crude oil participating in blending optimization comprise linear and nonlinear blending rules; x is the formula of each blending component oil, and X is ═ X₁,x₂…x_m]，x_i≥0；x_min，iAnd x_max，iRespectively the minimum value and the maximum value, x, of the formula of the blending component oil i_min，i≥0，x_max，i≥0；λ_lWeight of property I, lambda, for crude oil participation optimization_l≥0；goal_lFor the optimal target value, good, of the first property of the crude oil to participate in the optimization_l≥0；rangeL_lFor the lower target limit of the first property for crude oil to participate in optimization, range U_lThe target upper limit for the first property of the crude oil to participate in the optimization.

The invention has the beneficial effects that:

the crude oil nonlinear optimization blending method based on partition initialization solves the problem that the optimization property target is easy to fall into a local extreme value or cause no solution in optimization when a nonlinear blending rule is adopted in the crude oil online blending process. The conditions that the atmospheric and vacuum distillation device is seriously corroded, the produced product is unqualified and the like due to the fact that the atmospheric and vacuum distillation device is trapped in a local extreme value during optimization or the optimization is not solved are avoided.

Drawings

FIG. 1 is a flow chart of a crude oil nonlinear optimization blending method based on expanding an initialization range.

Detailed Description

The invention is further described with reference to the accompanying drawings and examples.

The invention is actually implemented in a refinery and combines with a specific example to give a detailed calculation process and a specific operation flow. The enterprise mainly processes the crude oil such as Jenno, Darr blending, Bazier, sand heavy, sand medium, Aman and the like, and in order to reduce the production cost, the low-sulfur crude oil and the high-sulfur crude oil are mixed and processed. In order to reduce the corrosion degree of the atmospheric and vacuum distillation unit during safe production, the content values of the blended crude oleic acid, the sulfur and the like are controlled within the design parameter range of the atmospheric and vacuum distillation unit. Meanwhile, in order to improve the octane number of the gasoline produced by the secondary processing device, the octane number of the crude oil is also used as an optimized property item. The present embodiment is implemented on the premise of the technical solution of the present invention, but the scope of the present invention is not limited to the following embodiments.

As shown in fig. 1, a crude oil nonlinear optimization blending method based on an expanded initialization range mainly comprises the following steps:

I. and acquiring basic parameters of crude oil blending.

In combination with a specific example, assuming that the crude oil blended and blended in this batch is jiuno, daler mixed crude oil and crude oil in basala, the properties participating in optimization include sulfur content, acid number, cetane number and residual oil sulfur content, and the properties of each oil, i.e. optimization constraints, are shown in table 1:

TABLE 1 participating in blending crude Properties and optimizing constraints

The blending rule adopted by the octane number is a quadratic nonlinear regression model with the largest weight, and the octane number calculation is taken as an example, and the formula is as follows:

in the formula (1), f (x) is the octane number of the blended crude oil; m is the component number participating in crude oil blending; x is the number of_iIs the blending mass ratio of the component crude oil i; q_i,_jThe blending effect coefficient of blending component crude oil i and j, the size and the positive and negative of which reflect the nonlinear action between different crude oils, Q_i,_j≠Q_j,i，Q_i,i＝Q_j,j＝1；R_i,jAs the arithmetic mean of the research octane numbers of the blended component crude oils i and j, R_i,_j＝(R_i+R_j)/2。

The octane blending effect coefficients for the above jenno, daler blends and basalas are shown in the following table:

TABLE 2 RON blending Effect coefficients between crude oils

And II, obtaining a crude oil blending nonlinear optimization algorithm. The objective function of the crude oil nonlinear blending optimization algorithm is as follows:

s.t.∑x_i＝1,x_i≥0

x_min，i≤x_i≤x_max,i

i＝1,2,…m

rangeL_l≤f(X)_l≤rangeU_l (7)

in the formula (7), Φ (X) is an objective function; h represents the number of properties of the crude oil participating in the optimization, f (X)_lBlending rules adopted for the first property of the crude oil participating in blending optimization comprise linear and nonlinear blending rules; x is the formula of each blending component oil, and X is ═ X₁,x₂…x_m]，x_i≥0；x_min，iAnd x_max，iRespectively the minimum value and the maximum value, x, of the formula of the blending component oil i_min，i≥0，x_max，i≥0；λ_lWeight of property I, lambda, for crude oil participation optimization_l≥0；goal_lFor the optimal target value, good, of the first property of the crude oil to participate in the optimization_l≥0；rangeL_lFor the lower target limit of the first property for crude oil to participate in optimization, range U_lAn upper target limit for the first property of the crude oil to participate in the optimization; specific constraints are shown in table 1.

III, limiting the formula upper limit (x) of each blending component_max,i) And lower limit (x)_min,i) Both are multiplied by (1+ 10%) and (1-10%) to obtain the upper limit range [ x ] of the extended formula_maxl,i,x _maxu,i]Expanding the lower limit range of the formula [ x_minl,i,x _minu,i]Narrow the range of the formula [ x_minu,i,x_maxl,i]And expanding the formulation range [ x ]_minl，i,x_maxu，i]. As shown in table 3:

TABLE 3 Small recipe ranges for blended crudes

Based on the calculation procedures (a) to (d), 9 sets of recipe initial values were generated according to table 3:

a) selecting the upper limit range [ x ] of the formulation of the ith component_maxl，i,x_maxu，i]Within this range, the mass fraction value X is randomly generated_bg,i-1With the remaining m-1 components in their extended formulation range [ x ]_minl，j,x_maxu，j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass fraction formula Xbe_i，1；

b) Selecting the lower limit range [ x ] of the formulation of the ith component_minl，i,x_minu，i]Within this range, a value X is randomly generated_bg,i-2With the remaining m-1 components in their extended formulation range [ x ]_minl，j,x_maxu，j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass fraction formula Xbe_i，2；

c) Selecting a reduced formula range [ x ] of the ith component_minu，i,x_maxl，i]Within this range, a value X is randomly generated_bg,i-3With the remaining m-1 components in their extended formulation range [ x ]_minl,j,x_maxu,j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass fraction formula Xbe_i，3；

d) Repeat steps a, b, c for all components to yield 3m initial set of parts by mass formulation Xbe.

The calculation results are shown in table 4:

TABLE 4 Small recipe Range Table for blended crudes

And V, sequentially selecting the initial mass ratio from the table 4, and calling a crude oil nonlinear blending optimization algorithm to obtain a feasible solution.

The final optimized feasible solution is shown in table 5.

Table 5 gives the feasible solutions and blended crude properties

From table 5, a total of 4 possible solutions were generated.

The four solutions are processed according to the objective function min

And the optimal solution of the final selection is [0.0151,0.3286,0.6563 ] calculated by preferentially meeting the octane number requirement]。

Claims (4)

1. A crude oil nonlinear optimization blending method based on expanded initialization range is characterized by comprising the following steps: expanding the formula range of m components participating in blending crude oil to generate an expanded formula range, generating 3m groups of initial mass part ratios in the expanded formula upper limit range, the expanded formula lower limit range, the reduced formula range and the expanded formula range, further performing optimization solution to obtain an optimized formula of the component crude oil, calculating the feasibility of each optimized formula solution, forming a set of optimized formula solutions from feasible optimized formula solutions of the component crude oil, and selecting an optimal group of solutions as a final formula of the component crude oil according to the size of a target function value in a crude oil nonlinear blending optimization algorithm; the method comprises the following specific steps:

I. obtaining basic parameters of crude oil blending, including the number m of components participating in blending, the properties of the crude oil, and the formula range XR of each component_iWhere i is 1,2 … m, optimizing property range constraints;

II, establishing a crude oil nonlinear blending optimization equation;

III, limiting the formula upper limit x of each blending component_max,iAnd a lower limit x_min,iBoth are multiplied by (1 +%) and (1-%) to obtain the upper limit range [ x ] of the expanded formula_maxl,i,x_maxu,i]Expanding the lower limit range of the formula [ x_minl,i,x_minu,i]Reducing the formulation range [ x_minu,i,x_maxl,i]And expand the formula range [ x_minl，i,x_maxu，i]The upper and lower limits of the formula for blending component i are scaled to a value between 0 and 10]To (c) to (d);

in the range of expanding upper limit of formula [ x ]_maxl,i,x_maxu,i]Expanding the lower limit range of the formula [ x_minl,i,x_minu,i]Reducing the formulation range [ x_minu,i,x_maxl,i]And expand the formula range [ x_minl，i,x_maxu，i]Generating 3m groups of initial mass part ratios Xbe; the calculation process is as follows:

a) selecting the upper limit range [ x ] of the formulation of the ith component_maxl，i,x_maxu，i]Within this range, the mass fraction value X is randomly generated_bg,i-1With the remaining m-1 components in their extended formulation range [ x ]_minl，j,x_maxu，j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass ratio of Xbe_i，1；

b) Selecting the lower limit range [ x ] of the formulation of the ith component_minl，i,x_minu，i]Within this range, a value X is randomly generated_bg,i-2With the remaining m-1 components in their extended formulation range [ x ]_minl，j,x_maxu，j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass ratio of Xbe_i，2；

c) Selecting a reduced formula range [ x ] of the ith component_minu，i,x_maxl，i]Within this range, a value X is randomly generated_bg,i-3With the remaining m-1 components in their extended formulation range [ x ]_minl,j,x_maxu,j]Each also generates 1 random number in the random number table for m-1 random number X_bg,jUsing the m random numbers as a group to optimize the initial mass ratio of Xbe_i，3；

d) Repeating the steps a, b and c on all the components to generate 3m groups of initial mass part ratios Xbe;

v, selecting a t-th group of new initial mass part ratios Xbe from the step IV_tInitially, t is 1;

VI, Xbe, with other blending basic parameters unchanged_tCalling a crude oil nonlinear blending optimization algorithm to solve an optimized formula XS of the component crude oil for initial mass part ratio_t；

VII, judging the optimized formula XS of the component crude oil solved in the step VI_tIf the solution is a feasible solution, entering the step VIII if the solution is a feasible solution, otherwise, directly entering the step IX;

VIII. mixing XS_tRecord to X_FSEntering the step IX;

if t is t +1, entering a step V when t < 3m, and otherwise, entering a step X;

x, ending the feasible solution process of the optimized formula of the crude oil;

XI after the process of obtaining a feasible solution is finished, compare X_FSThe optimal formula feasible solutions of all the crude oil components are selected as the final formula according to the value of the objective function in the crude oil nonlinear blending optimization algorithm.

2. The method as claimed in claim 1, wherein the calculation of properties of blending optimization during the blending optimization of crude oil is partially or totally based on nonlinear blending rules, wherein the octane number is calculated by a quadratic nonlinear regression model:

in the formula (1), f (x) is the octane number of the blended crude oil; m is the component number participating in crude oil blending; x is the number of_iIs the blending mass ratio of the component crude oil i; q_i,jThe blending effect coefficient of blending component crude oil i and j, the size and the positive and negative of which reflect the nonlinear action between different crude oils, Q_i,j≠Q_j,i，Q_i,i＝Q_j,j＝1；R_i,jAs the arithmetic mean of the research octane numbers of the blended component crude oils i and j, R_i,j＝(R_i+R_j)/2。

3. The method as claimed in claim 1, wherein the upper limit x of the formulation of each blending component is set as_max,iAnd a lower limit x_min,iBoth are multiplied by (1 +%) and (1-%) to obtain the upper limit range [ x ] of the expanded formula_maxl,i,x_maxu,i]And expanding the lower limit range of the formula [ x_minl,i,x_minu,i]The calculation process is as follows:

x_maxl，i＝x_max,i×(1-％) (3)

x_maxu，i＝x_max,i×(1+％) (4)

x_minl，i＝x_min,i×(1-％) (5)

x_minu，i＝x_min,i×(1+％) (6)

in formulae (2) to (6), XR_iFor blending the formula range of component i, x_max,iIs the maximum value of the formulation, x, of blending component i_min,iThe minimum value of the formula of the blending component i, m is the component number participating in crude oil blending, and the upper and lower limits of the formula of the blending component i are magnified and reduced and are between 0 and 10]In the meantime.

4. The method of claim 1, wherein the objective function of the nonlinear blending optimization algorithm for crude oil is as follows:

in the formula (7), Φ (X) is an objective function; h represents the number of properties of the crude oil participating in the optimization, f (X)_lBlending rules adopted for the first property of the crude oil participating in blending optimization comprise linear and nonlinear blending rules; x is the formula of each blending component oil, and X is ═ X₁,x₂…x_m]，x_i≥0；x_min，iAnd x_max，iRespectively the minimum value and the maximum value, x, of the formula of the blending component oil i_min，i≥0，x_max，i≥0；λ_lWeight of property I, lambda, for crude oil participation optimization_l≥0；goal_lFor the optimal target value, good, of the first property of the crude oil to participate in the optimization_l≥0；rangeL_lFor the lower target limit of the first property for crude oil to participate in optimization, range U_lThe target upper limit for the first property of the crude oil to participate in the optimization.

Images