Multi-refinery production plan optimization method, device and equipment and storage medium

Abstract

The invention discloses a multi-refinery production plan optimization method and device, electronic equipment and a storage medium. The method comprises the following steps: obtaining production element constraints according to production element parameters of multiple refineries; constructing a mixed integer nonlinear programming model according to the production element constraints, the decision variables and a preset objective function; wherein the decision variables comprise a plurality of integer variables; iteratively executing the following steps on the mixed integer nonlinear programming model until all branches are processed: determining and solving a next to-be-solved branch according to the sorting and assignment sequence of a plurality of integer variables, and updating the optimal solution of the mixed integer nonlinear programming model according to the optimal solution of the branch; and taking the optimal solution of the mixed integer nonlinear programming model as the production plan of the multiple refineries. According to the branch strategy based on the multi-refinery business characteristics, the branches are determined and solved according to the logistics circulation sequence and the start-stop priority among the devices, the convergence speed of the model can be remarkably increased, and the solving speed can be increased.

Description

Technical Field

[0001] This application relates to the technical field of production plan optimization, and particularly to a multi-refinery production plan optimization method, device, electronic device and storage medium. Background Art

[0002] Comprehensive petrochemical companies usually have multiple refineries. The crude oil resources allocated to multiple refineries need to be overallocated, and the products such as gasoline and diesel produced by multiple refineries also need to be overallocated to the market. Therefore, it is necessary to optimize the overall production plan for the allocation of crude oil among multiple refineries, the load of each refinery device, and the product structure.

[0003] However, since petroleum refining involves multiple complex refining processes and products, and refineries have hundreds of devices and thousands of materials, and there is a material mixing process in production, a large number of physical property calculations need to be performed on the mixed materials, which are expressed as non-convex bilinear type equalities or inequalities. Therefore, when jointly optimizing multiple refineries, the accumulation of non-linear constraints of each refinery will increase the difficulty of solving the model exponentially. In addition, due to the minimum threshold limit for the operating load of refinery devices, in some extreme cases, affected by factors such as insufficient upstream crude oil supply (such as natural disasters causing oilfield production reduction, pipeline maintenance causing a decrease in crude oil transportation volume, etc.) or a decrease in downstream product demand, it is also necessary to arrange for some devices of some refineries to stop production periodically. The start / stop of the device needs to be characterized by 0-1 variables, which makes the multi-refinery production plan optimization problem present in the form of large-scale mixed integer non-linear.

[0004] In related technologies, the solutions to the optimization problem mainly include Lagrangian relaxation, convex relaxation, and reconstruction linearization based on relaxation, distributed recursion, Benders decomposition based on linear programming, and commercial global optimization solvers. However, the above methods have at least the following problems in practical applications: The start / stop of devices is involved in the application scenario, which will introduce integer decision variables and cause a sharp increase in the difficulty of solving. When jointly optimizing multiple refineries, it will also cause difficulties in solving due to the increase in the scale of decision variables, with low solving efficiency. Moreover, in the face of complex scenarios, the convergence is slow and it is difficult to find a feasible solution, and it does not have good scenario applicability. Summary of the Invention

[0005] This application provides a multi-refinery production plan optimization method, device, electronic device and storage medium. Based on the branch strategy of the material flow sequence and start / stop assignment sequence in the production device, mixed integer non-linear programming (MINLP) is solved, so that the solving process can converge quickly and give high-quality solutions, significantly improving the model solving efficiency.

[0006] This application provides a multi-refinery production plan optimization method, including,

[0007] Obtain production factor constraints according to the production factor parameters of multiple refineries;

[0008] Construct a mixed-integer non-linear programming model according to the production factor constraints, decision variables, and a preset objective function; wherein, the decision variables include multiple integer variables;

[0009] Iteratively execute the following steps on the mixed-integer non-linear programming model until all branches are processed:

[0010] Determine the next branch to be solved and solve it according to the sorting and assignment order of the multiple integer variables, and update the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch;

[0011] Take the optimal solution of the mixed-integer non-linear programming model as the production plan of the multiple refineries;

[0012] Wherein, each of the integer variables indicates the start-stop state of a device in the multiple refineries; the sorting of the multiple integer variables indicates the transfer order of materials between the devices corresponding to the multiple integer variables.

[0013] This application also provides an optimization device for the production plan of multiple refineries, including,

[0014] A model constraint determination module, configured to obtain production factor constraints according to the production factor parameters of multiple refineries;

[0015] A model construction module, configured to construct a mixed-integer non-linear programming model according to the production factor constraints, decision variables, and a preset objective function; wherein, the decision variables include multiple integer variables;

[0016] A solving module, configured to iteratively execute the following steps on the mixed-integer non-linear programming model until all branches are processed: determine the next branch to be solved and solve it according to the sorting and assignment order of the multiple integer variables, and update the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch;

[0017] The solving module is further configured to take the optimal solution of the mixed-integer non-linear programming model as the production plan of the multiple refineries;

[0018] Wherein, each of the integer variables indicates the start-stop state of a device in the multiple refineries; the sorting of the multiple integer variables indicates the transfer order of materials between the devices corresponding to the multiple integer variables.

[0019] This application also provides an electronic device, including:

[0020] One or more processors;

[0021] A storage device for storing one or more programs,

[0022] When the one or more programs are executed by the one or more processors, the one or more processors implement the multi-refinery production plan optimization method as described in any embodiment of the present disclosure.

[0023] The present application also provides a computer storage medium storing a computer program, wherein the computer program is configured to execute the multi-refinery production plan optimization method as described in any embodiment of the present disclosure when running.

[0024] Compared with the related art, the solution of the present application is based on a solution method for mixed integer nonlinear programming (MINLP) of successive linear programming (SLP), and proposes a branching strategy based on the business characteristics of multi-refineries. According to the coupling order and start-stop priority of the devices, branches are determined and solved, which can significantly improve the model convergence speed, increase the solution speed, and effectively determine the multi-refinery production plan.

[0025] Other features and advantages of the present application will be described in the following specification, and some of them will become obvious from the specification, or be understood by implementing the present application. Other advantages of the present application can be realized and obtained through the solutions described in the specification and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings are used to provide an understanding of the technical solutions of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solutions of the present application, and do not constitute a limitation to the technical solutions of the present application.

[0027] Figure 1 A flowchart of a multi-refinery production plan optimization method provided for an embodiment of the present disclosure;

[0028] Figure 2 A schematic diagram of branch division provided for an embodiment of the present disclosure;

[0029] Figure 3 Another flowchart of a multi-refinery production plan optimization method provided for an embodiment of the present disclosure;

[0030] Figure 4 A structural diagram of a multi-refinery production plan optimization device provided for an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] This application describes multiple embodiments, but the description is exemplary rather than restrictive, and it will be apparent to those of ordinary skill in the art that there can be more embodiments and implementation solutions within the scope encompassed by the embodiments described in this application. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically restricted, any feature or element of any embodiment can be used in combination with any other feature or element in any other embodiment, or can replace any other feature or element in any other embodiment.

[0032] This application includes and contemplates combinations with features and elements known to those of ordinary skill in the art. The embodiments, features, and elements disclosed in this application can also be combined with any conventional features or elements to form unique inventive solutions defined by the claims. Any feature or element of any embodiment can also be combined with features or elements from other inventive solutions to form another unique inventive solution defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in this application can be implemented alone or in any suitable combination. Thus, the embodiments are not limited except as defined by the appended claims and their equivalents. Additionally, various modifications and changes can be made within the scope of the appended claims.

[0033] Furthermore, in describing representative embodiments, the specification may have presented the method and / or process as a particular sequence of steps. However, to the extent that the method or process does not depend on the particular order of the steps described herein, the method or process should not be limited to the particular order of steps described. As will be understood by those of ordinary skill in the art, other step orders are possible. Therefore, the particular order of steps set forth in the specification should not be construed as a limitation on the claims. Additionally, the claims directed to the method and / or process should not be limited to performing their steps in the order written, as those skilled in the art can readily understand that these orders can vary and still remain within the spirit and scope of the embodiments of this application.

[0034] At present, comprehensive oil and petrochemical companies at home and abroad usually have the characteristics of integration of upstream and downstream, production, supply and marketing. The crude oil produced by the company itself or purchased through long-term trade is overall allocated by the headquarters to multiple refineries within the region for processing, and the products such as gasoline and diesel produced by each refinery are then overall sold by the headquarters. At the headquarters level, it is necessary to comprehensively consider the upstream crude oil supply, the production constraints of each refinery in the midstream, and the market demand in the downstream, and optimize the allocation of crude oil among multiple refineries, the load of each refinery's devices, and the product structure, etc. overall, based on the principle of maximizing the overall efficiency of the company and the smooth operation of each link in the industrial chain. There are many factors involved in the above process, and it is very difficult to formulate an optimized and reasonable plan by manual means. It is necessary to adopt operational research optimization technology to establish an optimization model for the production plan of multiple refineries to provide support for the company's production and operation decisions.

[0035] The present embodiment uses the methods of operations research to establish an optimization model for the production plan of multiple refineries, and overall optimizes the allocation of crude oil among multiple refineries, the load of each refinery's devices, and the product plan. By using a mixed-integer nonlinear optimization algorithm based on continuous linear programming, it efficiently solves the optimization problem of multiple refineries with continuous variables by means of continuous linear programming, and then after embedding it into the branch and bound solution framework of integer programming, it can complete the decision-making with integers for the start and stop of devices, and further provide an optimized and reasonable production and operation decision-making plan for comprehensive oil and petrochemical companies, including the allocation plan of crude oil among multiple refineries, the start and stop plan of each refinery's devices, the production and processing plan of each refinery, and the product sales plan, etc., to improve the overall efficiency of the company.

[0036] In addition, in the actual application scenario, in the face of the mixed-flow problem involved in the refinery production process (which has been proven to be an NP-Hard problem) and the increased difficulty of solving caused by integer decision-making and complex large-scale decision-making brought about by multiple refineries, the multi-refinery production plan optimization method of the present application embodiment, based on the mixed-integer nonlinear programming (MINLP) solution method of continuous linear programming (SLP), proposes a branch strategy based on business characteristics, enabling the solution process to converge quickly and give high-quality solutions. And, combined with the actual scenario modeling of oil enterprises and relying on real cases for experiments and analysis, the experimental results show that the proposed algorithm has achieved good results.

[0037] The present disclosure embodiment provides a multi-refinery production plan optimization method, as Figure 1 shown, including,

[0038] Step 110, obtaining production factor constraints according to the production factor parameters of multiple refineries;

[0039] Step 120, constructing a mixed-integer nonlinear programming model according to the production factor constraints, decision variables, and a preset objective function; wherein, the decision variables include multiple integer variables;

[0040] Step 130: Iteratively execute the following steps on the mixed-integer non-linear programming model until all branches are processed:

[0041] Determine the next branch to be solved and solve it according to the sorting and assignment order of the multiple integer variables, and update the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch;

[0042] Step 140: Use the optimal solution of the mixed-integer non-linear programming model as the production plan of the multi-refinery;

[0043] Wherein, each of the integer variables indicates the start-stop state of a device in the multi-refinery; the sorting of the multiple integer variables indicates the transfer order of materials between the devices corresponding to the multiple integer variables.

[0044] In some exemplary embodiments, the sorting of the integer variables corresponding to the devices into which the materials in the multi-refinery first flow is prior to the sorting of the integer variables corresponding to the devices into which the materials then flow;

[0045] The assignment sorting indicating that the corresponding device is in the start state is prior to the assignment sorting indicating that the corresponding device is in the stop state;

[0046] The determining of the next branch to be solved according to the sorting and assignment order of the multiple integer variables includes:

[0047] Taking the sorting of the multiple integer variables as the node order and the assignment order of each integer variable as the left and right branch order, branch the mixed-integer non-linear programming model to determine the next branch to be solved.

[0048] For example, if 1 indicates that the device is in the start state and 0 indicates that the device is in the stop state, the sorting of the assignment of 1 is prior to the assignment of 0.

[0049] For another example, the multi-refinery includes an atmospheric distillation unit c1, a vacuum distillation unit v1, and a secondary unit u1, and the material transfer order is: atmospheric distillation unit - vacuum distillation unit - secondary unit. Then the sorting of the integer variable corresponding to the atmospheric distillation unit c1 is prior to the sorting of the integer variable corresponding to the vacuum distillation unit v1, the sorting of the integer variable corresponding to the vacuum distillation unit v1 is prior to the sorting of the integer variable corresponding to the secondary unit u1, the assignment sorting of the integer variable corresponding to the atmospheric distillation unit c1 with an assignment of 1 is prior to the assignment of 0, the assignment sorting of the integer variable corresponding to the vacuum distillation unit v1 with an assignment of 1 is prior to the assignment of 0, and the assignment sorting of the integer variable corresponding to the secondary unit u1 with an assignment of 1 is prior to the assignment of 0.

[0050] The atmospheric distillation unit c1, the vacuum distillation unit v1, and the secondary unit u1 correspond to three integer variables. Each integer variable includes two assignments: 1 and 0. Taking the sorting of these three integer variables as the node order and the assignment order as the left and right branch order, in the case where no pruning has been performed, the optimal solution of the model is selected from the branch optimal solutions of the following branches: c1(1)-v1(1)-u1(1), c1(1)-v1(1)-u1(0), c1(1)-v1(0)-u1(1), c1(1)-v1(0)-u1(0), c1(0)-v1(1)-u1(1), c1(0)-v1(1)-u1(0), c1(0)-v1(0)-u1(1), c1(0)-v1(0)-u1(0).

[0051] According to the above sorting and assignment order, the determined branches are as Figure 2 shown. The sorting determines the node order, that is, the node level, and the assignment order determines the left and right. The 8 branches correspond to 8 leaf nodes, and the leaf node order determined by the preorder traversal is the solution order of the 8 branches.

[0052] It should be noted that the above example only shows the situation of 3 units to clearly show the branch solution order of the embodiment of the present application. In the actual production scenario, multiple refineries may include more units, and the coupling order is set according to the corresponding units. More examples are not discussed one by one here.

[0053] In some exemplary embodiments, each of the branches corresponds to a non-linear sub-problem;

[0054] Each of the branches to be solved corresponds to a non-linear sub-model;

[0055] Each of the branches to be solved is solved according to the following method:

[0056] According to the node position of the branch to be solved, the assignments of the integer variables corresponding to the branched nodes are fixed as the parameters of the non-linear sub-model corresponding to this branch;

[0057] The non-linear sub-model is solved by using the successive linear programming SLP method to obtain the optimal solution of this branch.

[0058] It can be understood that the embodiments of the present disclosure use the successive linear programming SLP method to solve each non-linear sub-model.

[0059] In some exemplary embodiments, the solving the non-linear sub-model by using the successive linear programming SLP method to obtain the optimal solution of this branch includes:

[0060] Obtaining the corresponding linear sub-model according to the non-linear sub-model, and iteratively solving the linear sub-model to obtain the optimal solution of this branch;

[0061] Among them, the linear sub-model corresponding to the first branch is obtained by performing a first-order Taylor expansion on the non-linear sub-model corresponding to the first branch; the linear sub-model corresponding to a non-first branch is obtained by updating the model parameters related to integer variables based on the linear sub-model corresponding to the first branch.

[0062] In some exemplary embodiments, after the linear sub-model corresponding to the first branch is constructed, it is saved in memory or on a hard disk, read from it during the solution of subsequent branches, and the relevant model parameters are updated correspondingly according to the assignment of integer variables in the current branch to obtain the linear sub-model corresponding to the current branch.

[0063] In some exemplary embodiments, for each branch, when using the SLP method to solve the non-linear sub-model corresponding to the branch, the solution result with all integer variables being 1 is used as the initial solution of the current branch.

[0064] In some exemplary embodiments, the determining of the next branch to be solved and solving according to the sorting and assignment order of the multiple integer variables further includes:

[0065] In the case where the current branch has no solution, pruning is performed on the sub-branches of the current branch.

[0066] It can be understood that pruning the sub-branches of the current branch means that the current branch does not need to be further branched and solved. Thus, the overall number of branches to be solved can be reduced.

[0067] For example, for the branched node c1, v1, c1 = 1, v1 = 0, the current branch is the branch of c1 = 1, v1 = 0, and the parameters of the corresponding non-linear sub-model are fixed as c1 = 1, v = 0 for solution. If there is no solution, then pruning is performed on all sub-branches below c1 = 1, v1 = 0, and no further division and solution will be performed on other sub-branches below c1 = 1, v1 = 0. After pruning, the next branch to be solved is the branch of c1 = 0.

[0068] For another example, for the branched node c1, v1, u1, c1 = 1, v1 = 1, u1 = 1, the current branch is the branch of c1 = 1, v1 = 1, u1 = 1, and the parameters of the corresponding non-linear sub-model are fixed as c1 = 1, v1 = 1, u1 = 1 for solution. If there is no solution, then pruning is performed on all sub-branches below c1 = 1, v1 = 1, u1 = 1. According to Figure 2 It can be seen that there are no more sub-branches below c1 = 1, v1 = 1, u1 = 1, so no pruning is needed, or it can be understood as pruning an empty sub-branch. Correspondingly, the next branch to be solved is the branch of c1 = 1, v1 = 1, u1 = 0.

[0069] For another example, for the branched node c1 where c1 = 0, and the current branch is the branch where c1 = 0, the parameters of the corresponding non-linear sub-model are fixed at c1 = 0 for solution. If there is no solution, pruning is performed on all sub-branches below c1 = 0, and no further division and solution are performed on other sub-branches below c1 = 0. After pruning, there are no other branches to be solved, and the branch solution of the entire model ends.

[0070] It can be seen that branching, solving, and pruning according to the branch strategy provided by the embodiment of the present application can significantly reduce the number of branches actually solved and improve the model solving efficiency.

[0071] In some exemplary embodiments, the method further includes:

[0072] Step 100, set the initial value of the optimal solution of the mixed-integer non-linear programming model to be empty.

[0073] In some exemplary embodiments, the initial value of the optimal solution of the mixed-integer non-linear programming model is empty.

[0074] In some exemplary embodiments, updating the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of the branch includes:

[0075] According to the objective function, when it is determined that the optimal solution of the branch is better than the current value of the optimal solution of the mixed-integer non-linear programming model and satisfies the model optimal solution update condition, update the optimal solution of the mixed-integer non-linear programming model with the optimal solution of the branch;

[0076] Wherein, the model optimal solution update condition includes: all the multiple integer variables have been fixedly assigned values.

[0077] In some exemplary embodiments, if the objective function is to maximize the total profit of the upstream and downstream of the industrial chain, then when it is determined that the objective function value calculated corresponding to the optimal solution of the branch is greater than the objective function value calculated corresponding to the current value of the optimal solution of the mixed-integer non-linear programming model and satisfies the model optimal solution update condition, update the optimal solution of the mixed-integer non-linear programming model with the optimal solution of the branch. It can be understood that substituting the updated optimal solution of the mixed-integer non-linear programming model into the objective function, the currently maximum function value is calculated.

[0078] It should be noted that according to different objective functions, the optimal solution can be determined accordingly, not limited to the aspects exemplified in the present disclosure.

[0079] In some exemplary embodiments, updating the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of the branch further includes:

[0080] According to the objective function, when it is determined that the optimal solution of this branch is not better than the current value of the optimal solution of the mixed-integer non-linear programming model, pruning is performed on the sub-branches of this branch.

[0081] In some exemplary embodiments, the updating of the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch further includes:

[0082] According to the objective function, when it is determined that the optimal solution of this branch is better than the current value of the optimal solution of the mixed-integer non-linear programming model but does not meet the model optimal solution update condition, according to the sorting and assignment order of the multiple integer variables, continue to branch the corresponding non-linear sub-model of this branch to determine the next branch to be solved.

[0083] In some exemplary embodiments, the objective function is to maximize the total profit of the upstream and downstream of the industrial chain.

[0084] In some exemplary embodiments, the total profit is expressed according to the following formula: Total profit = sales revenue of each refinery - procurement cost of each refinery + sales revenue of the headquarters - procurement cost of the headquarters - transportation cost of the headquarters.

[0085] In some exemplary embodiments, the device includes an atmospheric distillation unit, a vacuum distillation unit, and a secondary unit;

[0086] The flow sequence of the material is: from the atmospheric distillation unit to the vacuum distillation unit, and then to the secondary unit.

[0087] In some exemplary embodiments, the description of the model-related set symbols is as follows:

[0088] Symbol description

[0089]

[0090]

[0091] That is, nodes (including oil fields, refineries, sales markets, etc.), refineries, materials, devices, and physical property sets are respectively defined as N, R, M, E, and QUA. Corresponding to specific sub-sets, raw materials, products, headquarters materials (referring to materials included in the company's overall allocation, such as crude oil, gasoline, diesel, etc.), and inventory material sets are respectively defined as M BUY , M SEL , M Z , M S ; The in-device and out-device sets of a certain material m ∈ M are respectively defined as and The feed set of the device e ∈ E is defined as

[0092] Define the refinery product sales price, raw material procurement price, and unit material conversion rate as α r,e,m′,m ; Define the headquarters product sales price, raw material procurement price, and material freight as

[0093] Define the refinery product sales volume, raw material procurement volume, end - of - period inventory, material property value, and upper and lower limits of unit processing volume as Define the upper and lower limits of the headquarters product sales volume, raw material procurement volume, end - of - period inventory, and material transportation volume as

[0094] According to the above production factor parameters, obtain the following production factor constraints:

[0095] The specific constraint function for the output material volume of the refinery unit is:

[0096]

[0097] The specific constraint function for the material property value of the refinery is:

[0098]

[0099] The specific constraint function for the unit processing volume of the refinery is:

[0100]

[0101] The specific constraint function for the material balance of the refinery is:

[0102]

[0103] The specific constraint function for the material balance of the headquarters node is:

[0104]

[0105] The specific constraint function for the upper and lower limits of the refinery raw material procurement and product sales is:

[0106]

[0107]

[0108] The specific constraint function for the upper and lower limits of the refinery unit processing capacity is:

[0109]

[0110] The specific constraint function for the upper and lower limits of the refinery material end - of - period inventory is:

[0111]

[0112] The specific constraint functions for the upper and lower limits of the physical property values of refinery materials are as follows:

[0113]

[0114] The specific constraint functions for the upper and lower limits of the headquarters' raw material procurement and product sales are as follows:

[0115]

[0116]

[0117] The specific constraint functions for the upper and lower limits of the headquarters' material ending inventory are as follows:

[0118]

[0119] The specific constraint functions for the upper and lower limits of the headquarters' transportation volume are as follows:

[0120]

[0121] Considering the upper and lower limits of the processing capacity of refinery units during startup and shutdown, Constraint (8) is modified to obtain a new constraint:

[0122]

[0123] Among them, is a 0-1 integer variable corresponding to the startup and shutdown of the unit. When the variable takes the value of 1, the unit is open, that is, the unit starts to run; when the variable takes the value of 0, the unit is closed, that is, the unit stops running.

[0124] The preset objective function is to maximize the company's profit. Profit = sales revenue of each refinery - procurement cost of each refinery + sales revenue of the headquarters - procurement cost of the headquarters - transportation cost of the headquarters. The corresponding objective function (0) is as follows:

[0125]

[0126]

[0127] Constraint (1) is the conversion equation for the materials entering and leaving the units in the refinery. Constraint (2) is the physical property calculation equation for the materials during blending in the refinery. Constraint (3) is the processing capacity calculation equation for the units in the refinery. Constraint (4) is the balance equation for each material in the whole refinery. Constraint (5) is the balance equation for the headquarters' materials at each node. Constraints (6), (7), (9)-(15) are the range limitations of the model on the variables.

[0128] Decision variables, define and as the raw material procurement volume and product sales volume of the refinery, and are the initial and final inventory levels of refinery materials, is the feed or output quantity of the unit, Q r,m,q is the material property variable, is the processing quantity of the unit; and are the sales volume of headquarters products and the raw material procurement volume, and are the initial and final inventory levels of headquarters materials, is the material transportation volume from one node to another node, is the start / stop of unit e in refinery r (0-1 variable, integer variable).

[0129] Based on the above constraints, decision variables, and the pre-set objective function (0), the construction of a mixed-integer non-linear programming model based on multiple refineries is completed.

[0130] Specifically, in the objective function, P refers to profit, R represents the set of refineries, r represents any refinery, M represents the set of materials, m represents any material, MSEL represents the set of products produced by the refinery, MBUY represents the set of raw materials purchased by the refinery, M S represents the set of inventory materials, M Z represents the set of headquarters materials (referring to materials such as crude oil, gasoline, and diesel that are coordinated and distributed among multiple refineries by the company headquarters), z represents any headquarters material, N represents the set of production plan nodes (production plan nodes include oilfield nodes, refinery nodes, sales nodes, etc.), n, n2 represent any nodes in the production plan nodes, and (n, n2) represents from node n to node n2. Combining with the business, for materials that are coordinated and purchased and sold by the headquarters, the corresponding price in the refinery can be set to 0.

[0131] In the above constraint (1), represents the processing or output quantity of material m in unit e of refinery r, M (m,m′) represents the set of feed m' that can produce material m, α r,e,m′,m represents the conversion rate of using m' to produce m in unit e of refinery r, represents the feed quantity of material m' in unit e of refinery r, R represents the set of refineries, r represents any refinery, M represents the set of materials, m represents any material, E represents the set of units in the refinery, and e represents any unit, represents the set of units that produce material m.

[0132] In the above constraint (2), QUA (abbreviated as Q) represents physical property, Q r,m,q represents the q physical property value of material m in refinery r, Q r,m′,qRepresents the q physical property value of material m' in refinery r. Represents the output of material m in unit e of refinery r. Represents the feed amount of material m' in unit e of refinery r. R represents the set of refineries, r represents any refinery, M represents the set of materials in the refinery, m represents any material, E represents the set of units in the refinery, and e represents any unit. Represents the set of units that produce material m.

[0133] In the above constraint (3), Represents the processing volume of the refinery's unit. Represents the feed amount of material m in unit e of refinery r. Represents the set of materials entering unit e. R represents the set of refineries, r represents any refinery, m represents any material, E represents the set of units in the refinery, and e represents any unit.

[0134] In the above constraint (4), Represents the initial inventory of refinery materials. Represents the final inventory of refinery materials. Represents the set of units that produce material m. Represents the set of units that process material m. Represents the inflow (outflow) of material m in unit e of refinery r. R represents the set of refineries, r represents any refinery, M represents the set of materials, and m represents any material.

[0135] In the above constraint (5), Represents the quantity of material z purchased by node n. Represents the quantity of material z transported from node n to node n2. Represents the initial inventory of material z in node n. Represents the quantity of material z sold by refinery r. Represents the final inventory of material z in node n. N represents the set of nodes, n represents any node in the production plan, M Z Represents the materials of the headquarters, and z represents any material of the headquarters.

[0136] In the above constraint (6), Represents the lower limit value of the purchase quantity of material m in refinery r. Represents the upper limit value of the purchase quantity of material m in refinery r. Represents the quantity of material m purchased by refinery r. R represents the set of refineries, r represents any refinery, M BUY Represents the set of raw materials purchased by the refinery, and m represents any raw material.

[0137] In the above constraint (7), represents the lower limit value of the sales volume of product m in refinery r, represents the upper limit value of the sales volume of product m in refinery r, represents the quantity of product m sold by refinery r. R represents the set of refineries, r represents any refinery, and M SEL represents the set of products sold, and m represents any product.

[0138] In the above constraint (8), represents the lower limit value of the processing capacity of unit e in refinery r, represents the upper limit value of the processing capacity of unit e in refinery r, represents the processing volume of unit e in refinery r. R represents the set of refineries, r represents any refinery, E represents the set of units, and e represents any unit.

[0139] In the above constraint (9), represents the lower limit value of the end - of - period inventory of material m in refinery r, represents the upper limit value of the end - of - period inventory of material m in refinery r, represents the end - of - period inventory of material m in refinery r. R represents the set of refineries, r represents any refinery, and M S represents the set of inventory materials, and m represents any inventory material.

[0140] In the above constraint (10), represents the lower limit value of the q - physical property of material m in refinery r, represents the upper limit value of the q - physical property of material m in refinery r, Q r,m,q represents the q - physical property value of material m in refinery r. R represents the set of refineries, r represents any refinery, QUA represents the set of physical properties, and q represents any physical property.

[0141] In the above constraint (11), represents the lower limit value of the purchase quantity of material z at node n, represents the upper limit value of the purchase quantity of material z at node n, represents the purchase quantity of material m at node n. N represents the set of nodes, n represents any node in the production plan, and M Z represents the set of materials at the headquarters, and z represents any material at the headquarters.

[0142] In the above constraint (12), represents the lower limit value of the sales volume of product z at node n, represents the upper limit value of the sales volume of product z at node n, represents the sales volume of product m at node n. N represents the set of nodes, n represents any node in the production plan, and MZ It represents the material set of the headquarters, and z represents any material in the headquarters.

[0143] In the above constraint (13), represents the lower limit value of the ending inventory of material z in node n, represents the upper limit value of the ending inventory of material z in node n, represents the ending inventory of material z in node n, N represents the set of nodes, n represents any node in the production plan, and z represents the intersection of the materials in the headquarters and the inventory materials.

[0144] In the above constraint (14), represents the lower limit value of the transportation volume from node n to n2, represents the upper limit value of the transportation volume from node n to n2, represents the transportation volume of material z from node n to n2, N represents the set of nodes, and n, n2 represent any nodes in the production plan scheme.

[0145] In the above constraint (15), combined with the specific scenario and business, considering the start-stop decision of the device, constraint (8) can be converted into constraint (15), where represents the lower limit value of the processing capacity of refinery r, unit e, represents the upper limit value of the processing capacity of refinery r, unit e, represents the 0-1 integer variable of refinery r, unit e. When is 1, it means unit e is in operation. When is 0, it means unit e is shut down. represents the processing volume of refinery r, unit e. R represents the set of refineries, r represents any refinery, E represents the set of units, and e represents any unit.

[0146] The embodiment of the present disclosure also provides a method for optimizing the production plan of multiple refineries. As Figure 3 shown, it includes

[0147] Step 310, constructing a mixed-integer nonlinear programming model (also called the original problem model) and initializing it;

[0148] Step 320, determining whether there are still unsolved branches. If not, execute 340. If so, execute 330.

[0149] Step 330, determining the next unsolved branch to be solved;

[0150] Step 340, determining whether the current value of the optimal solution of the original problem model is empty. If it is empty, the original problem model has no solution. If it is not empty, output the current optimal solution;

[0151] Step 350, solve the branch to be solved;

[0152] Step 360, determine whether there is no solution; if there is no solution, execute Step 370, if there is a solution, execute Step 380;

[0153] Step 370, prune the branch, and after pruning, return to Step 320 to continue iteration;

[0154] Step 380, determine whether the optimal solution of this branch is better than the current value of the optimal solution of the original problem model; if so, execute 370, if not, execute 370;

[0155] Step 390, determine whether all integer variables of the current branch have been fixedly assigned values; if so, execute Step 3100, if not, return to Step 320 to continue iteration;

[0156] Step 3100, update the current value of the optimal solution of the original problem model.

[0157] Among them, the initial value of the optimal solution of the mixed-integer non-linear programming model in Step 310 is empty;

[0158] After fixing the assignment of the integer variable corresponding to the branched node as the parameter of the non-linear sub-model corresponding to this branch in Step 350, solve the sub-model.

[0159] For the constructed mixed-integer non-linear programming model, the inputs of the model include integer variables, integer constraints, linear constraints, and bilinear constraints. Each branch is regarded as a sub-problem, corresponding to a non-linear sub-model. According to the branch strategy based on the sorting and assignment order of integer variables proposed in the embodiments of the present disclosure, branch division and solution are performed. After fixing some or all integer variables for each non-linear sub-model in the order of nodes, solve them to achieve pre-pruning of some integer variables to reduce the number of branches of the exponential order. Since branch and bound is an exponential complexity algorithm, reducing the computational complexity of the SLP algorithm when solving sub-problems can improve the overall efficiency of the solution framework. Therefore, in this paper, when implementing the branch and bound algorithm framework, models are established for the original problem and sub-problems respectively, and only variable information is updated in each iteration to reduce the time for re-modeling sub-problems of each branch.

[0160] In each branch solution, the SLP algorithm uses two models, namely a non-linear sub-model with integer variables fixed as parameters and a linear sub-model, where the linear sub-model is constructed by performing a first-order Taylor expansion on the non-linear sub-model.

[0161] In some exemplary embodiments, the construction of the linear sub-model includes: storing the linear sub-model constructed by the first branch into the memory, and when solving the subsequent branches, updating the sub-model parameters related to the integer variables according to the stored linear sub-model, and obtaining the linear sub-model of the current branch, thereby realizing the rapid generation of a new linear sub-model. Compared with some implementable schemes, in which the linear sub-model is re-constructed from the corresponding nonlinear sub-model in each branch, the first construction proposed here and the subsequent update of the linear sub-model parameters can greatly reduce the time loss of the first-order expansion and linearization process of the nonlinear sub-model for each branch iteration, and only the assignment process is performed, which reduces the overall time complexity and has a relatively obvious execution speed advantage in the actual operation process.

[0162] It can be known that each time the SLP algorithm is used to solve a branch in the branch and bound framework, an initial solution is required. In order to quickly obtain a better initial solution, in some exemplary embodiments, for each branch, a strategy of using the solution result with all integer variables as 1 is adopted as the initial solution, and there is no need to find a separate initial point (initial solution) for the SLP calculation of each branch, which saves calculation time and improves the efficiency of model solving.

[0163] In the disclosed embodiment, continuous linear programming is embedded in the branch and bound framework, and the integer variables are fixed as model parameters at each integer node branch. The advantage of this is that the integer variable decision-making ability of branch and bound and the good convergence of continuous linear programming in nonlinear sub-problems are fully utilized, which greatly reduces the redundant decision-making of integer variables. The branch and bound algorithm based on continuous linear programming adopts a general integer branch framework.

[0164] In some exemplary embodiments, solving a branch includes: solving sub-problems corresponding to each branch using a sequential linear programming (SLP) method.

[0165] In some exemplary embodiments, the parameters of the sub-problem model further include an iteration index parameter, a trust region parameter, a penalty parameter, an iteration parameter, a scaling parameter, and an allowable value parameter.

[0166] Exemplarily, there are three preset discrimination parameters in the range of 0 to 1, specifically 0<ρ0<ρ1<ρ2<1. After the three preset discrimination parameters are sorted, they are divided into four discrimination intervals, specifically [0, ρ0], [ρ0, ρ1], [ρ1, ρ2], [ρ2, 1].

[0167] It can be understood that the first improvement data may be the actual improvement amount calculated:

[0168] Δp k =M ori (x k )-M ori (xk +d k );

[0169] The second improved data can be the calculated estimated improvement amount:

[0170] Δl k =M ori (x k )-M sub (x k ,d k );

[0171] Where Δp k represents the actual improvement amount, M ori represents the non-linear sub-model, k represents the iteration index, x k represents the solution of the non-linear sub-model corresponding to the iteration index, d k represents the preset differential variable, Δl k represents the estimated improvement amount, M sub represents a linear sub-model.

[0172] Based on the actual improvement amount and the estimated improvement amount, the improvement amount ratio can be obtained as follows:

[0173] r k =Δp k / Δl k ;

[0174] r k represents the improvement amount ratio. When the improvement amount ratio is in the interval [0, ρ0], the following assignment process is performed:

[0175] Δ k+1 ←δ1Δ k ;

[0176] δ1 represents the first trust region scaling parameter, and the value range of δ1 is between 0 and 1. Δ k represents the change in the function in the linear sub-model when the preset differential variable is d k . k represents the number. Assign the first trust region scaling parameter and the change in the function in the k-th linear sub-model to the change in the function in the (k + 1)-th linear sub-model.

[0177] It can be seen that for branch solution, first degenerate the non-linear sub-model corresponding to the branch into a series of linear models, and the above-mentioned k-th and (k + 1)-th linear sub-models are both one of the series of linear models.

[0178] When the improvement amount ratio is in the interval [ρ0, ρ1], the following assignment process is performed:

[0179] Δ k+1 =δ1Δ k ;

[0180] Δ k+1 ← max{Δ k+1 , Δ′};

[0181] k ← k + 1;

[0182] δ1 represents the first trust region scaling parameter, and the value range of δ1 is between 0 and 1. First, assign the change in the function in the k-th linear submodel and the first trust region scaling parameter to the change in the function in the (k + 1)-th linear submodel. Δ′ represents the minimum allowable value of the trust region, and Δ′ > 0. Compare the change in the function in the (k + 1)-th linear submodel with the minimum allowable value of the trust region, and update the maximum value of the two to the change in the function in the (k + 1)-th linear submodel, replacing the k-th moment with the (k + 1)-th moment.

[0183] When the improvement ratio is in the interval [ρ1, ρ2], perform the following assignment process:

[0184] Δ k+1 = Δ k ;

[0185] Assign the change in the function in the k-th linear submodel to the change in the function in the (k + 1)-th linear submodel.

[0186] When the improvement ratio is in the interval [ρ2, 1], perform the following assignment process:

[0187] Δ k+1 = δ2Δ k ;

[0188] Δ k+1 ← max{Δ k+1 , Δ′};

[0189] k ← k + 1;

[0190] δ2 represents the second trust region scaling parameter, and δ2 > 1. First, assign the change in the function in the k-th linear submodel and the second trust region scaling parameter to the change in the function in the (k + 1)-th linear submodel. Δ′ represents the minimum allowable value of the trust region, and Δ′ > 0. Compare the change in the function in the (k + 1)-th linear submodel with the minimum allowable value of the trust region, and update the maximum value of the two to the change in the function in the (k + 1)-th linear submodel, replacing the k-th moment with the (k + 1)-th moment.

[0191] Based on the description in the algorithm flow above, the solution idea is:

[0192] 1) Use an approximate linear submodel with a first-order Taylor expansion to make decisions on the moving direction and step size of the decision variables;

[0193] 2) The movement area is limited by the trust region method and updated according to the solution results of each approximate linear sub-model.

[0194] 3) The current trust region and solution are evaluated based on the objective function values and constraint violation performances of the two models, and the constraint violation value is gradually reduced and the objective function value is increased to search for the most likely optimal solution (branch optimal solution).

[0195] Taking the production scenarios of multiple refineries of an oil company as an example, the algorithm effect is experimentally verified. The processor of the experimental environment is an Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz 2.40GHz, and the memory is an 8GB computer. The Gurobi solver version 9.5.2 is used to solve the linear programming model and the mixed integer programming model. Two test cases are selected for the experiment, and the scale of the test cases is described by the business information and model information they contain, as summarized in Tables 1 and 2.

[0196] Table 1 Parameter information of two test cases

[0197]

[0198] Test case 1 includes 3 refineries, each refinery has more than 2 atmospheric and vacuum distillation units and multiple secondary processing units, and each of these 3 refineries has one atmospheric and vacuum distillation unit processing the same kind of crude oil. Test case 2 includes 5 refineries, each refinery has more than 2 atmospheric and vacuum distillation units and multiple secondary processing units producing a kind of gasoline.

[0199] It should be noted that the parameters of the production plan of the multiple refineries according to the embodiments of the present invention include relevant variables, parameters, specific parameters, etc. required in the model.

[0200] Table 2 Parameter information of the models of two test cases

[0201] Example Variable Constraint Integer variable Nonlinear constraint 1 18212 8614 78 3070 2 28933 13556 96 4764

[0202] Since the test cases used in the experiments of this embodiment are multiple refinery models, the solution scale is large, and due to the addition of integer decision variables, there is currently no Benchmark that can be used for evaluation. For general-purpose solution methods, Baron, as the fastest global optimization solver, cannot give a feasible solution reference within a reasonable time. Therefore, in this embodiment, a solution scheme of generating a feasible solution by turning off some devices according to business experience is selected for comparison with the proposed algorithm. The specific operation is as follows: According to business experience, the devices to be turned off are manually selected, the integer variables are preset, and the SLP method is used to solve a continuous variable problem to obtain a feasible solution scheme. We use the objective function value as the evaluation index and compare the method proposed in this embodiment with the results of the feasible solution scheme obtained based on business experience.

[0203] It should be noted that in the following table, "-" indicates that the solution time exceeds the set 3600s, meaning there is no solution.

[0204] Table 3 Comparison Information of Algorithm Effects

[0205]

[0206] Example 1 includes a total of 3 refineries. Each refinery has more than 2 atmospheric and vacuum distillation units and multiple secondary processing units. Each of the 3 refineries has an atmospheric and vacuum distillation unit to process the same crude oil MC. In a certain month, due to the production reduction of the upstream oilfield, the supply of crude oil MC cannot enable all the units of the 3 refineries to start operation simultaneously. It is necessary to apply the model to calculate the optimal allocation plan of crude oil MC and the unit startup plan of each refinery in this scenario.

[0207] Example 2 includes a total of 5 refineries. Each refinery has more than two atmospheric and vacuum distillation units and multiple secondary processing units. In a certain month, due to the decline in downstream gasoline consumption, the production capacity of the refineries is greater than the sales capacity, and it is impossible to arrange for all the units of the 5 refineries to start operation simultaneously. It is necessary to apply the model to calculate the most reasonable refinery unit startup plan and the product production plan in this scenario.

[0208] Among them, the impact of the number of refineries on the model scale mainly lies in the fact that there will be more couplings among multiple refineries. Moreover, due to the different business characteristics of each refinery, the types and quantities of units in each refinery are different, and the corresponding raw materials and products also vary. Therefore, the impact of the model scale on the solution is not a simple linear change. The number of variables is mainly related to refineries, materials, physical properties, and units. Among them, the number of physical properties directly determines the number of non-linear constraints, thereby affecting the difficulty of problem-solving. The number of integer variables directly determines the calculation scale of branch and bound. When there are more units that need to be decided to start or stop in the problem, the scale of integer variables is larger, and the calculation difficulty of branch and bound often increases accordingly.

[0209] In Example 1, the Baron solver adopted in the related technology has no solution. Compared with the manual method + SLP, the method of the embodiment of the present invention improves the objective function value by 3%. In Example 2, compared with the manual method + SLP, the method of the embodiment of the present invention improves the objective function value by 5%. Compared with the manual method + SLP, the method of the embodiment of the present invention can further increase the objective function related to the benefit, thereby ensuring the production plan benefit.

[0210] Table 4 Comparison Information of the Solution Results of Example 1

[0211]

[0212] Table 5 Comparison Information of the Solution Results of Example 2

[0213]

[0214] Based on the above, the method of the embodiment of the present invention can shut down more equipment and ensure greater benefits while ensuring production needs, thereby improving the management efficiency and operation efficiency of oil-related enterprises.

[0215] Table 6 Comparison of branch solution times of different branch strategies

[0216]

[0217]

[0218] Among them, c represents the normal pressure device, v represents the pressure reduction device, and u represents the secondary processing device. Using different setting orders for the three devices will result in different number of model solutions and ultimately different total solution times.

[0219] In the experiment in Table 6, the depth-first search direction of branch-and-bound is used by default, and the branch value is first branched to 1 and then branched to 0. The convergence time of other branches and value sequences is greater than 3600 seconds. In order to avoid the influence of the search direction of branch-and-bound on the solution, the experiment also compares the depth-first and breadth-first branch search directions in the branch-and-bound framework. The results given in Table 6 are for the depth-first search direction, and the breadth-first search cannot converge within 3600 seconds.

[0220] It can be seen from the three branching orders that branching the atmospheric pressure device c and the pressure reducing device v first often converges faster. After the secondary device is advanced, the larger-scale 2 example did not converge within 3600 seconds. Placing the atmospheric pressure device before the pressure reducing device also has certain advantages. This phenomenon is due to the fact that in actual refineries, the atmospheric pressure and pressure reducing devices that need to be decided to start and stop are often relatively small, while the number of secondary devices is relatively large. Taking Example 1 as an example, the number of atmospheric pressure, pressure reducing and secondary devices that need to be decided is 7, 4, and 67 respectively.

[0221] The order of the integer variable corresponding to the secondary device u is too high, which will result in an attempt to shut down some of the normal pressure or pressure reducing devices when the secondary device is fully open, causing the material to stop flowing in the production line composed of the normal pressure device, the pressure reducing device and the secondary device, and it is impossible to obtain an integer feasible solution.

[0222] For the business scenarios of the embodiments of the present invention, the order of branch values ​​in the breadth-first search has no effect on the final effect. From the perspective of the branch order of the device, since each device must first complete the 1-0 branch, it is equivalent to opening and closing each device once in business, which is out of the order of material flow in the actual refining production process, so a lot of time will be spent on the front devices.

[0223] It can be understood that branching the atmospheric distillation unit and the vacuum distillation unit is often fast. Bringing forward the secondary processing unit, for the multi-refinery production situation of Case 2, it is difficult to converge within 3600 seconds. The present invention does not limit the specific branching strategy. Under the condition of meeting the production requirements and computing power requirements, the branching strategy can be reasonably selected to ensure obtaining the optimal integer solution of the model and meeting the requirements of both depth-first and breadth-first search directions.

[0224] Therefore, based on the branching strategy of this embodiment, determine the node order and the branching value order, fix the integer variables for the non-linear sub-problem, and realize the pre-pruning of some integer variables and reduce the number of branching times of the exponential order of magnitude. Since the branch and bound is an exponential complexity algorithm, adding a branching strategy can greatly reduce the number of branching times and accelerate the convergence speed of integer decision-making. Reducing the computational complexity of the SLP algorithm when solving sub-problems can improve the overall efficiency of the solution framework. Therefore, in this embodiment, when implementing the branch and bound algorithm framework, models are established for the original problem and the linear sub-problem respectively, and only the variable information is updated in each iteration to reduce the time for re-modeling the sub-problem in each branch.

[0225] The embodiments of the present disclosure also provide a multi-refinery production plan optimization device, as Figure 4 shown, including,

[0226] A model constraint determination module 410, configured to obtain production factor constraints according to the production factor parameters of the multi-refinery;

[0227] A model construction module 420, configured to construct a mixed-integer non-linear programming model according to the production factor constraints, decision variables, and a preset objective function; wherein, the decision variables include a plurality of integer variables;

[0228] A solution module 430, configured to iteratively execute the following steps on the mixed-integer non-linear programming model until all branches are processed: determine the next branch to be solved and solve it according to the sorting and assignment order of the plurality of integer variables, and update the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch;

[0229] The solution module 430 is further configured to use the optimal solution of the mixed-integer non-linear programming model as the production plan of the multi-refinery;

[0230] Wherein, each of the integer variables indicates the start-stop state of a device in the multi-refinery; the sorting of the plurality of integer variables indicates the transfer order of materials between the devices corresponding to the plurality of integer variables.

[0231] The embodiments of the present disclosure also provide an electronic device, including:

[0232] One or more processors;

[0233] a storage device for storing one or more programs,

[0234] When the one or more programs are executed by the one or more processors, the one or more processors implement the multi-refinery production planning optimization method as described in any embodiment of the present disclosure.

[0235] An embodiment of the present disclosure further provides a computer storage medium, in which a computer program is stored, wherein the computer program is configured to execute the multi-refinery production plan optimization method as described in any embodiment of the present disclosure when running.

[0236] In order to solve the multi-refinery production plan optimization problem with integer variables and mixed flows, and realize the overall optimization of crude oil distribution among multiple refineries, the load of each refinery unit and the product plan, the production plan optimization method of the embodiment of the present invention proposes a mixed integer nonlinear programming (MINLP) solution method based on continuous linear programming (SLP). The branching strategy based on the material flow sequence and the start-stop assignment sequence in the production unit enables the solution process to converge quickly and give high-quality solutions. In addition, it combines the actual scenario modeling of oil companies, and relies on real cases for experiments and analysis. The experimental results show that the proposed algorithm has achieved good results. It has good scenario applicability and effectively solves the problems of slow convergence and difficulty in finding feasible solutions in complex scenarios.

[0237] The production plan optimization method of the embodiment of the present invention utilizes a mixed integer nonlinear optimization algorithm based on continuous linear programming, and uses continuous linear programming to efficiently solve the multi-refinery optimization problem of continuous variables. After being embedded in the branch and bound solution framework of integer programming, it can complete integer decisions on the start and shutdown of equipment, thereby providing a comprehensive petroleum and petrochemical company with optimized and reasonable production and operation decision-making plans, including crude oil distribution plans among multiple refineries, equipment start and shutdown plans for each refinery, production and processing plans for each refinery, and product sales plans, etc., to improve the overall benefits of the company.

[0238] Those of ordinary skill in the art will understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and appropriate combinations thereof. In the hardware implementation, the division of the functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be executed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically includes computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

Claims

1. A method for optimizing the production plan of multiple refineries, characterized in that, including obtaining production factor constraints according to the production factor parameters of multiple refineries constructing a mixed-integer nonlinear programming model according to the production factor constraints, decision variables, and a preset objective function; wherein the decision variables include multiple integer variables iteratively performing the following steps on the mixed-integer nonlinear programming model until all branches are processed determining the next branch to be solved and solving it according to the sorting and assignment order of the multiple integer variables, and updating the optimal solution of the mixed-integer nonlinear programming model according to the optimal solution of this branch using the optimal solution of the mixed-integer nonlinear programming model as the production plan of the multiple refineries wherein each of the integer variables indicates the start-stop state of a device in the multiple refineries; the sorting of the multiple integer variables indicates the transfer order of materials among the devices corresponding to the multiple integer variables 2. The method according to claim 1, characterized in that, the sorting of the integer variable corresponding to the device into which the material first flows in the multiple refineries precedes the sorting of the integer variable corresponding to the device into which the material then flows the assignment sorting indicating that the corresponding device is in the start state precedes the assignment sorting indicating that the corresponding device is in the stop state the determining the next branch to be solved according to the sorting and assignment order of the multiple integer variables includes branching the mixed-integer nonlinear programming model with the sorting of the multiple integer variables as the node order and the assignment order of each integer variable as the left and right branch order to determine the next branch to be solved 3. The method according to claim 1 or 2, characterized in that, each of the branches to be solved corresponds to a nonlinear sub-model each of the branches to be solved is solved according to the following method fixing the assignments of the integer variables corresponding to the branched nodes as the parameters of the nonlinear sub-model corresponding to this branch according to the node position where this branch to be solved is located solving the nonlinear sub-model using the continuous linear programming SLP method to obtain the optimal solution of this branch 4. The method according to claim 3, characterized in that, the solving the nonlinear sub-model using the continuous linear programming SLP method to obtain the optimal solution of this branch includes obtaining the corresponding linear sub-model according to the nonlinear sub-model and iteratively solving the linear sub-model to obtain the optimal solution of this branch wherein the linear sub-model corresponding to the first branch is obtained by performing a first-order Taylor expansion on the nonlinear sub-model corresponding to the first branch; the linear sub-model corresponding to a non-first branch is obtained by updating the model parameters related to the integer variables based on the linear sub-model corresponding to the first branch 5. The method according to claim 3, characterized in that, the determining the next branch to be solved and solving it according to the sorting and assignment order of the multiple integer variables further includes pruning the sub-branches of the current branch when the current branch has no solution 6. The method according to claim 3, characterized in that, further including setting the initial value of the optimal solution of the mixed-integer nonlinear programming model to be empty the updating the optimal solution of the mixed-integer nonlinear programming model according to the optimal solution of this branch includes updating the optimal solution of the mixed-integer nonlinear programming model with the optimal solution of this branch when it is determined that the optimal solution of this branch is better than the current value of the optimal solution of the mixed-integer nonlinear programming model and meets the model optimal solution update condition according to the objective function Among them, the optimal solution update condition of the model includes: all the multiple integer variables have been fixedly assigned values.

7. The method according to claim 6, characterized in that, Updating the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch further includes: According to the objective function, when it is determined that the optimal solution of this branch is not better than the current value of the optimal solution of the mixed-integer non-linear programming model, pruning the sub-branches of this branch.

8. The method according to claim 6, characterized in that, Updating the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch further includes: According to the objective function, when it is determined that the optimal solution of this branch is better than the current value of the optimal solution of the mixed-integer non-linear programming model but does not meet the optimal solution update condition of the model, continue to branch the non-linear sub-model corresponding to this branch according to the sorting and assignment order of the multiple integer variables to determine the next branch to be solved.

9. The method according to claim 1 or 2, characterized in that, The objective function is to maximize the total profit of the upstream and downstream of the industrial chain.

10. The method according to claim 9, characterized in that, The total profit is expressed according to the following formula: Total profit = sales revenue of each refinery - procurement cost of each refinery + sales revenue of the headquarters - procurement cost of the headquarters - transportation cost of the headquarters.

11. The method according to claim 1 or 2, characterized in that, The device includes an atmospheric distillation unit, a vacuum distillation unit and a secondary unit; The flow sequence of the materials is: from the atmospheric distillation unit to the vacuum distillation unit, and then to the secondary unit.

12. An optimization device for multi-refinery production planning, characterized in that, Including, A model constraint determination module, configured to obtain production factor constraints according to the production factor parameters of multiple refineries; A model construction module, configured to construct a mixed-integer non-linear programming model according to the production factor constraints, decision variables and a preset objective function; among them, the decision variables include multiple integer variables; A solution module, configured to iteratively execute the following steps on the mixed-integer non-linear programming model until all branches are processed: determine the next branch to be solved and solve it according to the sorting and assignment order of the multiple integer variables, and update the optimal solution of the mixed-integer non-linear programming model according to the optimal solution of this branch; The solution module is further configured to use the optimal solution of the mixed-integer non-linear programming model as the production plan of the multiple refineries; Among them, each of the integer variables indicates the start-stop state of a device in the multiple refineries; the sorting of the multiple integer variables indicates the flow sequence of materials among the devices corresponding to the multiple integer variables.

13. An electronic device, characterized in that, Including: One or more processors; A storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the multi-refinery production plan optimization method according to any one of claims 1-11.

14. A computer storage medium, characterized in that, A computer program is stored in the storage medium, wherein the computer program is configured to execute the multi-refinery production plan optimization method according to any one of claims 1-11 when running.

Images