A crude oil blending optimization method for complex properties and scheduling inventory

Abstract

The present invention discloses a crude oil blending optimization method for complex properties and scheduling inventory. The method takes into account linear and nonlinear crude oil properties and crude oil scheduling inventory, and designs an improved sparrow search optimization to achieve crude oil blending, which can cope with complex blending situations with many nonlinear property constraints and crude oil scheduling restrictions. The method is improved on the basis of the traditional sparrow search algorithm, so that it can better adapt to nonlinear constraints and further improve the global search performance, improve its optimization ability, and avoid falling into a local optimal situation after premature convergence. In addition, the method promotes a better coordination between the scheduling link and the blending link by taking the crude oil scheduling inventory into consideration, promotes the integration of crude oil scheduling and blending, and further improves the economic benefits of refining enterprises.

Description

Technical Field

[0001] The invention relates to the field of crude oil blending in the oil refining industry, and in particular to a method for realizing crude oil blending by taking into account linear and nonlinear crude oil properties and crude oil scheduling inventory, and designing an improved sparrow search optimization. Background Art

[0002] In order to obtain as much low-priced crude oil as possible in the international crude oil market and reduce the company's crude oil procurement costs, current refining companies tend to purchase crude oil in multiple batches and multiple oil types. However, this method has led to an increase in the number of oil types and a decrease in the amount of oil from a single oil type, making crude oil blending and scheduling management more complicated. In addition, the current crude oil blending is developing in a more refined direction, and the control of properties, especially the blending of some nonlinear properties, is increasingly required. Therefore, in the actual production process of refining companies, crude oil blending needs to consider not only linear crude oil properties, but also nonlinear crude oil properties, and also needs to take into account the coordination of blending and upstream and downstream to achieve long-term stable production operation.

[0003] As a new type of intelligent optimization algorithm, the sparrow search optimization method (SSA) has good performance in solving optimization problems with severe nonlinearity and has been successfully applied in many fields in recent years. However, how to make the properties of the blended oil as stable as possible while meeting the needs of crude oil scheduling and considering the many nonlinear key properties of crude oil, so as to adapt to the crude oil scheduling situation and ensure that the constant pressure reduction unit does not stop production and the product properties are stable, which puts higher requirements on the corresponding intelligent optimization methods. The traditional sparrow search algorithm has the problems of easy premature convergence and falling into local optimality, weak optimization ability and difficulty in introducing complex nonlinear constraints. How to overcome the above shortcomings has become a key issue in the optimization of crude oil blending under the current complex crude oil properties and scheduling conditions. Summary of the invention

[0004] The purpose of the present invention is to provide a method for implementing crude oil blending by taking into account complex nonlinear crude oil properties and crude oil scheduling inventory, and designing an improved sparrow search algorithm. The method can quickly and accurately calculate the crude oil blending ratio under the conditions of many nonlinear property constraints and crude oil scheduling restrictions.

[0005] The present invention has the following steps:

[0006] (1) Initializing crude oil blending parameters, including inputting the number, properties and inventory of the type of oil to be processed, and initializing a sparrow population with n individuals, where the position of the individual is represented by X = [X1, X2, ..., X d ] represents a blending ratio scheme using d types of crude oil as raw materials. Each individual in the sparrow population satisfies Randomly initialize under the condition of max, producer ratio a1, sentinel ratio a2; set the initial value of iteration number l to l=1, the number of producers PD=n·a1, the number of predators LD=n·(1-a1), and the number of sentinels SD=n·a2.

[0007] (2) Calculate the linear weighted and nonlinear weighted blended crude oil properties respectively, where:

[0008] The linear weighted crude oil property indicators include sulfur content, acid value, naphtha yield and normal alkanes. The corresponding blended crude oil property calculation formula is:

[0009]

[0010] Where U θ is the θth linear weighted property index of blended crude oil, is the θth linear weighted property index of component crude oil k, S NQ is a collection of linearly weighted crude oil properties;

[0011] The nonlinear weighted crude oil property indicators include density, pour point and 50°C viscosity, and the corresponding blended crude oil property calculation formulas are:

[0012] ① Blended crude oil density SG:

[0013]

[0014] Among them, SG k represents the density of crude oil component k, X k It represents the mass blending ratio of component crude oil k;

[0015] ② Pour point P of blended crude oil:

[0016]

[0017] Among them, p k It represents the pour point of component crude oil k;

[0018] ③50℃ viscosity CS of blended crude oil:

[0019]

[0020] Where VF is calculated as follows:

[0021]

[0022] Among them CS k It represents the 50℃ viscosity of crude oil component k.

[0023] (3) Determine the upper and lower limits of crude oil blending ratio and the upper and lower limits of crude oil properties, and establish the basic crude oil blending optimization objective function

[0024]

[0025] where η k represents the price of crude oil component k, λ θ represents the weight of the θth property index of crude oil, U θ * The expected value X of the θth property index of the blended crude oil k is the blending ratio of crude oil k.

[0026] (4) According to the following formula, the crude oil scheduling inventory is introduced into the crude oil blending optimization model to achieve the adaptation of crude oil blending to scheduling, and the improved crude oil blending optimization objective function is obtained:

[0027]

[0028] Where C is the reserve influencing factor, V k,t It refers to the real-time cumulative reserves of component crude oil k in the tank area of ​​the factory at the current time t, and d is the total number of component crude oils in a blending plan.

[0029] (5) Determine the fitness function fit(X k ) and calculate the individual fitness value of the sparrow, where:

[0030]

[0031] Where η k represents the price of component oil k; θ represents the weight of the θth property index of crude oil; U θ Indicates the θth property index value of blended crude oil, U θ * represents the expected value of the θth property index of blended crude oil; NA represents the number of crude oil indexes, which are divided into linear indexes: sulfur content, acid value, naphtha yield and normal alkanes, and nonlinear indexes: density, pour point and 50℃ viscosity; C is the reserve influencing factor. The larger the C value, the greater the influence of reserves on the blending ratio of oil products; V k,l is the reserves of crude oil of the kth component at the lth iteration; δ is the penalty factor. The larger the δ value is, the higher the priority of considering this constraint in the solution. Changing the equality constraints of boundary conditions into inequality constraints is beneficial to boundary condition control.

[0032] (6) Perform position update: take the first PD sparrow individuals as producers and the last LD sparrow individuals as predators, and update the positions of producers and predators according to the current position, global optimal position and global worst position of the sparrows in the previous iteration; take the first SD sparrow individuals as sentinels, and update the sentinel position according to the optimal position of the producer in the previous iteration, the global worst position of the sparrow population and the position of the sentinel individuals; obtain the new position of all individuals in the current iteration round, and determine whether the new position is better than the previous position according to the fitness function. If the new position is better, use the new position to update the individual position, otherwise do not update the individual position;

[0033] (7) According to the following formula, a Gaussian random walk strategy that tends to be optimal is introduced to prevent the traditional sparrow search algorithm from falling into the local optimum too early:

[0034]

[0035] in When the number of iterations is l, the individual The new individuals obtained after random walk are is the global optimal position at iteration number l, is the position of the u-th sparrow corresponding to crude oil k at the time of iteration l, γ is the weight used to adjust the randomness of the random walk, ξ1 and ξ2 are random numbers that obey the 0-1 uniform distribution, The mean is Gaussian distributed random numbers with standard deviation σ.

[0036] (8) The degraded individual mutation strategy is introduced to improve the optimization ability of the traditional sparrow search algorithm. First, two random real numbers β and ε with a value range of [0,1] are generated according to the order of individual fitness values ​​in the sparrow population from small to large, and the following operation is performed on the sparrow individuals that meet the degraded individual standards:

[0037]

[0038] in, represents a degraded individual that has undergone mutation. is an individual u1 randomly selected from all individuals in the population, is the individual u2 (u2≠u1) randomly selected from all individuals in the population.

[0039] (9) Introducing a boundary control strategy to improve the solving ability of the traditional sparrow search algorithm after introducing complex nonlinear crude oil blending constraints, calculate the discriminant △ 1,θ =U θ,min -U θ and 2,θ =U θ -Uθ,max , if △ 1,θ ≤0 and △ 2,θ ≤0, Then calculate the fitness value of the individual, otherwise set the fitness value of the individual to +∞, where U θ Indicates the θth property index of blended crude oil, U θ,max , U θ,min It represents the upper and lower limits of the θth property of the blended crude oil, S NA Represents the set {1…NA} of crude oil properties.

[0040] (10) Update the individual ranking, obtain the global optimal individual, and determine whether the maximum number of iterations ω has been reached max If it is not reached, then l = l + 1 and return to (5), otherwise the entire optimization process ends and the result is the ωth max The crude oil blending ratio represented by the global optimal individual in the iteration.

[0041] Beneficial effects:

[0042] The present invention discloses a crude oil blending optimization method for complex properties and scheduling inventory. The method takes into account linear and nonlinear crude oil properties and crude oil scheduling inventory, and designs an improved sparrow search optimization to achieve crude oil blending, which can cope with complex blending situations with many nonlinear property constraints and crude oil scheduling restrictions. The method is improved on the basis of the traditional sparrow search algorithm, so that it can better adapt to nonlinear constraints and further improve the global search performance, improve its optimization ability, and avoid falling into a local optimal situation after premature convergence. In addition, by taking the crude oil scheduling inventory into consideration, it promotes better coordination between the scheduling link and the blending link, helps to promote the integration of crude oil scheduling and blending, and further improves the economic benefits of refining enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of a crude oil blending optimization method for complex properties and scheduling inventory in the present invention;

[0044] Figure 2 It is the iterative convergence curve diagram of crude oil blending using the traditional sparrow search algorithm;

[0045] Figure 3 It is the iterative convergence curve diagram of crude oil blending using the improved sparrow search algorithm. DETAILED DESCRIPTION

[0046] The following is a detailed description of the implementation effect of the method in crude oil blending under complex properties and scheduling conditions with reference to the accompanying drawings and specific examples. This implementation case is implemented based on the technical solution of the present invention, but the protection scope of the present invention is not limited to the following embodiments.

[0047] This embodiment is an example of crude oil blending in a certain refining enterprise. In this embodiment, five kinds of conventional crude oil are used for blending optimization, namely Basra Light, Basra Heavy, Kuwait, Fordis and Castile. In this embodiment, the atmospheric and vacuum unit needs to select three component crude oils from the five kinds of conventional crude oils for blending according to the process requirements to obtain the best crude oil blending formula.

[0048] First, determine the crude oil involved in the blending and its properties. According to the process requirements, the crude oil blending in this embodiment needs to consider the sulfur content, acid value, naphtha yield, normal alkanes, density, pour point and 50°C viscosity of the seven crude oil properties. From the crude oil property library provided by the refining enterprise, the properties and price data of five kinds of common crude oils can be obtained as shown in Table 1:

[0049] Table 1 Prediction results of samples in batches A and B

[0050]

[0051] When the scheduling start time of this embodiment is 60 hours, the real-time cumulative inventory of each oil type in the storage tank area of ​​the refinery is shown in Table 2:

[0052] Table 2 Real-time cumulative inventory of various oil types in the tank area of ​​the factory (t = 60h)

[0053]

[0054] According to the production and processing route of the atmospheric and vacuum unit, the feed blending crude oil property limits and target values ​​are determined as shown in Table 3:

[0055] Table 3 Limitations and target values ​​of blended crude oil properties

[0056]

[0057] Next, the crude oil blending optimization model is constructed. This embodiment uses a penalty function to convert the equality constraint into a non-equality constraint, which requires that the penalty function needs to be large enough to approximate the equality constraint. In this embodiment, the penalty factor δ is set to 1.0×10 8 . In addition, each crude oil property has its corresponding optimization weight λ. In the actual application process of the refining enterprise, the process personnel can set the optimization weight in the objective function according to the requirements of the on-site process for the importance of the crude oil properties. The crude oil properties with higher optimization weights λ will be met first. This embodiment pays more attention to the adaptability of the blending formula to the inventory than any property or price. Therefore, the crude oil reserve impact factor in the objective function is set to C=100, the optimization weight λ of each crude oil property is 10, and the optimization weight of the crude oil price is η=20.

[0058] Set the relevant parameters of sparrow optimization SSA. Specifically: the number of sparrow population n = 100, the maximum number of iterations ωmax =500, the producer ratio is a1=0.7, the predator ratio is 1-a1=0.3, and the sentinel ratio is a2=0.2, then the number of producers PD=na1=70, the number of predators LD=n(1-a1)=30, the number of sentinels SD=na2=20, and the random walk randomness weight γ in the improved SSA is 1.

[0059] In this case, the feed of the atmospheric and vacuum unit requires that the blended crude oil is a mixture of 3 component crude oils. There are 5 types of refining crude oils, so there are 10 component oil combinations for crude oil blending. This case enumerates all component oil combinations and optimizes them one by one. For the unselected oil types in the enumerated component oil combinations, the upper and lower limits of the blending ratio are set to [0,0], and for the selected oil types, the upper and lower limits of the blending ratio are set to [0.01,0.99]. The blending ratios of Basra Light, Kuwait, Fordis, Basra Heavy and Castile are x1, x2, x3, x4, and x5 respectively.

[0060] The traditional SSA and improved SSA are used to optimize the crude oil blending formula for all crude oil components, and the solution results are shown in Tables 4 and 5. The objective function is to minimize the target, so the smaller the fitness value, the better.

[0061] Table 4. Solution results of traditional SSA algorithm

[0062]

[0063] Table 5 Solution results of improved SSA algorithm

[0064]

[0065] *Note: Inf in the table means that there is no solution for the crude oil combination of this component in crude oil blending optimization.

[0066] From Table 4, we can see that the optimal crude oil blending formula in the traditional SSA solution is [x1:x2:x3:x4:x5= 0.1056:0.3469:0.0000:0.0000:0.5475]; from Table 5, we can see that the optimal crude oil blending formula in the improved SSA solution is [x1:x2:x3:x4:x5=0.0292:0.5434:0.0000:0.0000:0.4274]. By comparing with Table 2, we can see that the Basra light oil corresponding to x4 in the current plant is 58664.96t, the Kuwait oil corresponding to x2 is 195506.48t, and the Castile oil corresponding to x5 is 84761.54t. It can be seen that no matter it is the traditional SSA or the improved SSA, after considering the scheduling inventory, the results of the solution are in line with the principle that the blending ratio of the oil type with the larger inventory in the warehouse should be adjusted as high as possible.

[0067] The optimization performance of traditional SSA and improved SSA is compared and analyzed. The comparison indicators are shown in Table 6:

[0068] Table 6 Comparison of the optimal blending formula of traditional SSA and improved SSA

[0069]

[0070] It can be seen from Table 6 that the optimization results of traditional SSA and improved SSA are close to the target values, but the optimal blending formula obtained by improved SSA is better than that of traditional SSA in terms of relative density, sulfur content, naphtha yield, normal alkanes, pour point and viscosity. In addition, the optimal blending formula obtained by improved SSA has lower cost, which can further reduce the production cost of enterprises on the basis of meeting the feed requirements of refining enterprises for crude oil processing units.

[0071] Figure 2 and Figure 3 These are the iterative convergence curves of crude oil blending using traditional and improved SSA. Figure 2 and Figure 3 , the traditional SSA converged in the 5th iteration, while the improved SSA converged and stabilized after the 189th iteration. It can be seen that the traditional SSA converged prematurely, while the improved SSA has a better optimal solution search capability.

[0072] In summary, compared with the traditional crude oil blending optimization method, the method designed by the present invention has better optimal solution search capability and can avoid the phenomenon of falling into local optimum after premature convergence. This method can not only meet the complex nature requirements of the feed of the processing unit, but also take into account the scheduling situation, which is helpful to improve the economic benefits of the refining enterprise.

Claims

1. A crude oil blending optimization method for complex properties and scheduling inventory, characterized by Taking into account not only the linear and nonlinear crude oil properties but also the crude oil scheduling inventory, an improved sparrow search optimization method is designed to improve the crude oil blending optimization performance, including the following steps: (1) Initializing crude oil blending parameters; (2) Calculate linear weighted and nonlinear weighted crude oil properties respectively; the nonlinear weighted crude oil properties include density, pour point and 50°C viscosity, and the corresponding calculation formulas for the crude oil properties are: ① Blended crude oil density SG: Among them, SG k represents the density of the component crude oil k; ② Pour point P of blended crude oil: Among them, p k It represents the pour point of component crude oil k; ③50℃ viscosity CS of blended crude oil: Where VF is calculated as follows: Among them CS k It represents the 50℃ viscosity of the crude oil component k; (3) Determine the upper and lower limits of crude oil blending ratio and the upper and lower limits of crude oil properties, and establish the basic crude oil blending optimization objective function X k is the blending ratio of crude oil k; (4) Introducing crude oil scheduling inventory The improved crude oil blending optimization objective function is obtained Where C is the reserve influencing factor, V k,t It refers to the real-time cumulative reserves of component crude oil k in the tank area of ​​the plant at the current time t, and d is the total number of component crude oils in a blending plan; (5) Determine the fitness function of the sparrow search algorithm and calculate the fitness value of the individual sparrow; (6) Perform location update; (7) Introduce the optimal Gaussian random walk strategy according to the following formula: in When the number of iterations is l, the individual The new individual obtained after random walk, γ is the weight used to adjust the randomness of random walk, The mean is Gaussian random numbers with standard deviation σ, is the global optimal position at iteration number l, is the position of the u-th sparrow corresponding to crude oil k at the time of iteration number l, ξ1 and ξ2 are random numbers that obey the uniform distribution of 0-1; (8) Introduce the degraded individual mutation strategy. According to the order of individual fitness values ​​in the sparrow population from small to large, generate two different random real numbers β and ε with a value range of [0,1], and perform the following operation on the sparrow individuals that meet the degraded individual standards: in represents a degraded individual that has undergone mutation. is an individual u1 randomly selected from all individuals in the population, is the individual u2 randomly selected from all individuals in the population, u2≠u1; (9) Introduce the boundary control strategy and calculate the discriminant Δ 1,θ =U θ,min -U θ and Δ 2,θ =U θ -U θ,max , if Δ 1,θ ≤0 and Δ 2,θ ≤0, Then calculate the fitness value of the individual, otherwise set the fitness value of the individual to +∞, where U θ Indicates the θth property index of blended crude oil, U θ,max , U θ,min It represents the upper and lower limits of the θth property of the blended crude oil, S NA The set {1…NA} representing crude oil properties; (10) Update the individual ranking, obtain the global optimal individual, and determine whether the maximum number of iterations ω has been reached max If it is not reached, the number of iterations is l = l + 1 and returns to step (5), otherwise the entire optimization process ends and the result is the ωth max The crude oil blending ratio represented by the global optimal individual in the iteration.

2. The crude oil blending optimization method for complex properties and scheduling inventory according to claim 1 is characterized in that The optimization objective function of base crude oil blending in step (3) for: where η k represents the price of crude oil component k, λ θ represents the weight of the θth property index of crude oil, U θ * It represents the expected value of the θth property index of blended crude oil.

3. The crude oil blending optimization method for complex properties and scheduling inventory according to claim 1 is characterized in that The crude oil scheduling inventory V in step (4) k,t Provided by the on-site crude oil dispatching system.

4. The crude oil blending optimization method for complex properties and scheduling inventory according to claim 1 is characterized in that The fitness function fit(X k ) is calculated using the following formula: Where η k represents the price of crude oil component k, λ θ represents the weight of the θth property index of crude oil, U θ * V represents the expected value of the θth property index of blended crude oil, k,l is the reserves of crude oil of the kth component at the lth iteration, and δ is the penalty factor.

5. The crude oil blending optimization method for complex properties and scheduling inventory according to claim 1 is characterized in that The σ in step (7) is calculated as follows:

Images