CN117875477A - Production model optimization method suitable for complex refinery - Google Patents

Abstract

The invention discloses a production model optimization method suitable for a complex refinery, which comprises the following steps: obtaining a general petroleum refining plan optimization model; modeling and optimizing conversion are carried out based on the structure and characteristics of the petroleum production process; solving an optimal plan optimization scheme in a constraint model space by using a nonlinear programming method based on a branch-and-bound method; the method can improve the plan optimization estimation precision of petroleum refining, particularly can effectively prevent abnormal optimization jump and suboptimal solution in the aspect of processing a large-scale nonlinear problem of plan optimization, comprehensively considers constraint limits in the plan, improves calculation stability and accuracy, shortens solving time, and achieves the aim of improving production benefits.

Description

Production model optimization method suitable for complex refinery

Technical Field

The invention relates to the technical field of petrochemical production plan optimization, in particular to a production model optimization method suitable for complex refineries.

Background

Usually, the planning optimization is market driven, and the final objective of the planning optimization is to meet the market demand for products in an optimal manner under the constraints of raw materials and production operations. Because the types of production processes and devices will vary slightly, the problem of the instant bilinear property when the petroleum refining plan optimizes the property calculation, the blending involved may be three-linear, whereas for the factory, the non-linear constraints involved in the whole field are very high. However, in the prior art, for a large-scale plan optimization problem, an optimal scheme cannot be found, divergence may be solved or the solving time is too long, the adopted searching method is greatly dependent on an initial value problem, and the solving result is greatly dependent on the quality of the initial value.

For example, a "a refinery production plan optimizing method and apparatus" disclosed in chinese patent literature, its bulletin number: CN111598306B discloses a network topology model comprising constructing a refinery production process, and according to the network topology model and each process model, establishing a production plan optimization model with maximum economic benefit as a target and with processing capacity, material properties, market supply and demand and material balance as constraints; solving the production plan optimization model by adopting a queuing competition algorithm to obtain an optimal production method; however, in the scheme, because the process model is a nonlinear programming model, a production plan optimization model established based on the process model cannot be solved by adopting methods such as simultaneous equations and the like, and can only be solved by adopting a plurality of heuristic algorithms, so that the global convergence has a certain limitation, only local optimal solutions can be obtained, and the solution time is long on medium and large-scale problems.

Disclosure of Invention

In order to solve the problem that the optimization method in the prior art depends on the initial value, the invention provides a production model optimization method suitable for a complex refinery, improves the plan optimization estimation precision of petroleum refining, effectively prevents abnormal jump-out and suboptimal solution of optimization, and improves the calculation stability and accuracy.

In order to achieve the above object, the present invention provides the following technical solutions:

a production model optimization method suitable for complex refineries comprises the following steps,

obtaining a general petroleum refining plan optimization model;

modeling and optimizing conversion are carried out based on the structure and characteristics of the petroleum production process;

and solving the optimal plan optimization scheme in the constraint model space by using a nonlinear programming method based on a branch-and-bound method. By carrying out model conversion and pretreatment on a general petroleum refining plan optimization model, the structure of the model is simplified, variables or constraints are reduced, the scale and complexity of the problem are reduced, the problem is easier to solve, and the stability, accuracy and efficiency of the planning optimization in petroleum refining are improved.

Preferably, the optimizing conversion includes, performing variable analysis to determine an effective value range of the variable on a feasible region; performing constraint condition analysis, and screening redundant constraint and contradictory constraint according to linear correlation and nonlinear correlation among constraint conditions; and carrying out objective function analysis, and carrying out objective item balance on the contribution of the whole objective function according to each objective item. Preprocessing is achieved through variable analysis, constraint condition analysis and objective function analysis in optimization conversion, wherein the variable analysis verifies the range of variables, the constraint condition analysis verifies the reliability of constraint conditions, and the objective function analysis adjusts the accuracy of an objective function; finally, the aims of reducing solving difficulty and improving stability, accuracy and efficiency of plan optimization are fulfilled.

Preferably, the modeling includes obtaining structural parameters and characteristic parameters of the petroleum production process, wherein the structural parameters are content parameters of different component structures after the petroleum is processed, and the characteristic parameters are process parameters corresponding to different processing stages of the petroleum. Determining the change condition of a component structure in the petroleum processing process through structural parameters, and determining the process conditions of different stages in the petroleum processing process through characteristic parameters; the content parameters comprise real-time content parameters, historical content parameters and future content parameters, and the technological parameters comprise processing time, processing procedures and the like. Accurate oil refining data are obtained when the optimization of the petroleum production model is realized.

Preferably, in a general petroleum refining plan optimization model, basic data of refinery processing is acquired, processing indexes of the refinery are set as optimization constraints, and a production process flow chart is drawn. The basic data comprise device data, process data, raw material data, intermediate product data, product data and price data of refinery processing, different processing indexes are respectively set for the raw materials, the intermediate products and the products, and constraints respectively targeting the raw materials, the intermediate products and the products are determined according to the processing indexes of the raw materials, the intermediate products and the products. Process data for the raw materials, intermediates or products are determined by plotting a production process flow diagram. And the statistical visualization of the whole process of refinery processing is realized.

Preferably, the productivity estimation is performed after the drawing of the production process flow chart, including performing device decomposition on the production process, performing device modeling based on the product yield and the property, and performing the productivity estimation of the device yield. Dividing a production process flow chart into a plurality of sections of curves according to different devices, modeling each section of curve according to the yield and the properties of reactants in the devices to obtain a capacity estimation model, and determining the actual capacity of each device according to the capacity estimation model. Accurate capacity estimation of complex processes in refineries is achieved.

Preferably, after the capacity estimation is performed, a plurality of constrained plan constraint models are established, a plan optimization objective function is determined according to the plan constraint models, and the plan optimization objective function is the sum of optimization objectives of each plan constraint model. Based on the productivity estimation model, adding constraint taking the machining index as optimization, thereby establishing a plan constraint model, and combining the plan constraint model with a plan target to obtain a plan optimization objective function, so as to obtain the optimization objective function reflecting the total efficiency.

Preferably, in the solving based on the branch-and-bound method, different constraints are respectively bound to obtain a plurality of optimal solutions, and the plurality of optimal solutions are combined to be used as an optimal planning scheme. Determining a feasible region of each constraint, and establishing a search tree containing sub-problems, namely a subset of a search space; and (3) iteratively calculating the sub-problems in a feasible domain to obtain an optimal solution of the sub-problems, updating the optimal solution of the sub-problems into a search tree, and pruning the sub-problems when the sub-problems have no solution better than the existing optimal solution. Finally, all constraints form branches of the search tree, and all optimal solutions are combined to serve as an optimal planning scheme when unexplored problems are not found. The optimization of complicated steps, various raw materials and various intermediate products in the complex refinery is realized.

Preferably, the petroleum refining characteristic parameters are obtained, the petroleum refining characteristic parameters are cut after passing through each refining device, a plurality of main characteristic parameters and a plurality of auxiliary characteristic parameters are obtained through cutting, and an optimal solution is obtained for each main characteristic parameter and each auxiliary characteristic parameter through a branch and bound method. By cutting the petroleum refining characteristic parameters, the characteristic parameters are obtained in more detail and accuracy, so that the parameters of the optimization process change along with the refining process, the accuracy of the optimization process is not completely dependent on the quality of the initial value, abnormal jump-out and suboptimal solution of the optimization are effectively prevented, and the calculation stability and accuracy are improved.

The invention has the following advantages:

(1) The plan optimization estimation precision of petroleum refining can be improved, especially, abnormal jump-out and suboptimal solution of optimization can be effectively prevented on the aspect of processing a large-scale nonlinear problem of plan optimization, constraint restriction in the plan is comprehensively considered, calculation stability and accuracy are improved, solving time is shortened, and the aim of improving production benefits is fulfilled; (2) The model conversion and pretreatment are carried out on the general petroleum refining plan optimization model, so that the structure of the model is simplified, variables or constraints are reduced, the scale and complexity of the problem are reduced, the problem is easier to solve, and the stability, accuracy and efficiency of the planning optimization in petroleum refining are improved; (3) The optimization of complicated steps, various raw materials and various intermediate products in the complex refinery is realized, so that the accuracy of the optimization process does not depend on the quality of initial values completely.

Drawings

The drawings in the following description are merely exemplary and other implementations drawings may be derived from the drawings provided without inventive effort for a person of ordinary skill in the art.

FIG. 1 is a schematic diagram of the method steps in an embodiment.

FIG. 2 is a flow chart of a refinery in an embodiment.

In the figure:

1-an atmospheric and vacuum device; 2-catalytic cracking unit; 3-diesel oil blending device.

Detailed Description

All other embodiments obtained in the embodiments of the invention without making any inventive effort fall within the scope of the invention.

As shown in fig. 1-2, in a preferred embodiment, the present invention discloses a method for optimizing a production model for a complex refinery, comprising the steps of,

obtaining a general petroleum refining plan optimization model;

modeling and optimizing conversion are carried out based on the structure and characteristics of the petroleum production process;

and solving the optimal plan optimization scheme in the constraint model space by using a nonlinear programming method based on a branch-and-bound method.

When the method is used, a general bilinear or trilinear model in the petroleum refining process is firstly obtained, model conversion and pretreatment are carried out on the general bilinear or trilinear model, a global resource plan optimization model with a plurality of equality constraints and nonlinear terms is obtained, and iterative optimization is carried out through a branch-and-bound method to obtain an optimal plan scheme.

In other embodiments, the optimizing conversion includes performing variable analysis to determine an effective value range of the variable in the feasible region; performing constraint condition analysis, and screening redundant constraint and contradictory constraint according to linear correlation and nonlinear correlation among constraint conditions; and carrying out objective function analysis, and carrying out objective item balance on the contribution of the whole objective function according to each objective item. The preprocessing is realized through variable analysis, constraint condition analysis and objective function analysis in the optimization conversion, wherein the variable analysis verifies the range of the variable, the constraint condition analysis verifies the reliability of the constraint condition, and the objective function analysis adjusts the accuracy of the objective function.

In variable analysis, the type of each variable, including decision variables, state variables, or auxiliary variables, is first determined.

Decision variables are variables that need to be optimized, while state variables are system states or condition variables in the problem, auxiliary variables are typically used to simplify the model or introduce constraints.

For each variable, a range of values is determined, including an upper bound and a lower bound, the range of variables being used to determine the limits of the optimization space and the search direction.

The influence of each variable on the problem is analyzed, including determining the relationship between the variable and the objective function, constraint condition.

The roles of the variables in the constraints are analyzed to determine their limits on the feasible solutions.

Correlations between variables are determined, including linear correlations and non-linear correlations.

The interdependencies between variables are determined by calculating covariance, correlation coefficients or correlation graphs between the variables. To help determine which variables play an important role in the optimization process and whether the variables can be further simplified or combined.

The constraints to which each variable is subjected are analyzed, including equality constraints and inequality constraints.

Determining constraints on the variables can help determine the feasible region of the problem and determine the effective range of values of the variables over the feasible region.

In constraint analysis, the type of each constraint is first determined, including equality constraints and inequality constraints.

Equality constraints require that certain variables satisfy a particular equality relationship, while inequality constraints require that certain variables satisfy an inequality relationship.

Determining the form of the constraint includes analyzing the specific form of each constraint. Including linear constraints, non-linear constraints, convex constraints, or non-convex constraints.

Determining the form of the constraint helps determine whether a particular optimization technique needs to be applied to handle that type of constraint.

For each constraint, its upper and lower bounds, as well as other boundary conditions, are determined. Determining the constraints of the constraints helps to determine the range of feasible solution spaces and the search direction of the solver.

The influence of each constraint on the problem is determined, including determining the relationship between the constraint and the objective function, and their degree of restriction on the variables in the feasible domain.

Determining compatibility and relationships between constraint conditions, determining whether redundant or contradictory constraints exist, and how to interpret the redundant or contradictory constraints.

Linear and nonlinear correlations between constraints are determined to determine dependencies between constraints.

And the feasible solution space is expanded or the search direction is limited by relaxing or enhancing part of constraint conditions, so that the solution effect is improved.

In objective function analysis, the type of objective function is first determined, the problem is minimized or maximized, which will determine the objective of optimizing the problem.

Specific forms of the objective function are analyzed, including linear, nonlinear, convex, or non-convex. Determining the form of the objective function helps to determine the applicable solution and optimization techniques.

And decomposing the objective function, and analyzing each objective item in the objective function, wherein the objective item comprises linear combination of variables, an exponential function and a power function.

The contribution of each target term to the overall objective function is determined, thereby determining the role and importance of each target term in the optimization process. This helps to determine which target items have a dominant effect on the optimization results and to do special processing or weighting.

Determining the properties of the objective function, including convexity, microposity, continuity, facilitates selection of an appropriate optimization algorithm and solver, and determines constraints in the optimization process.

The objective function is optimized, including weighting, normalization, or regularization. By adjusting the form of the objective function or introducing penalty terms, the emphasis of the optimization problem can be changed or multiple objectives balanced.

In other embodiments, the modeling includes obtaining structural parameters and characteristic parameters of the petroleum production process, where the structural parameters are content parameters of different component structures after the petroleum is processed, and the characteristic parameters are process parameters corresponding to different processing stages of the petroleum. Determining the change condition of a component structure in the petroleum processing process through structural parameters, and determining the process conditions of different stages in the petroleum processing process through characteristic parameters; the content parameters comprise real-time content parameters, historical content parameters and future content parameters, and the technological parameters comprise processing time, processing procedures and the like. Accurate oil refining data are obtained when the optimization of the petroleum production model is realized.

When the device is used, reactants in the production process are monitored through the structural parameters and the characteristic parameters respectively, and when the structural parameters are changed, the reactants are represented to be changed; determining the change of reactants in the production process of a refinery according to the change of the structural parameters, and determining a piecewise curve composed of different reactants; meanwhile, the refining production process of various reactants is monitored according to the characteristic parameters.

Corresponding basic data are firstly obtained from a general petroleum refining plan optimization model, then classified and converted, and finally converted into a data form.

In other embodiments, basic data of refinery processing is obtained in a general petroleum refining plan optimization model, a processing index of the refinery is set as an optimization constraint, and a production process flow chart is drawn. The basic data comprise device data, process data, raw material data, intermediate product data, product data and price data of refinery processing, different processing indexes are respectively set for the raw materials, the intermediate products and the products, and constraints respectively targeting the raw materials, the intermediate products and the products are determined according to the processing indexes of the raw materials, the intermediate products and the products.

The constraints include device yield balance, device property balance, device processing capacity constraint, material property constraint, device fortification constraint and public works balance.

When in use, basic data such as devices, materials, properties, public works, process flows and the like are input; the basic data is divided into raw material related data, intermediate product related data and product related data.

Setting the processing capacity limit, material purchase and sale price, upper limit and lower limit of the device, and setting the feeding fortification and blending product quality index of the device; and respectively taking the raw materials, the intermediate products and the products as targets to determine the processing index.

Drawing a whole factory process flow chart, and perfecting property transfer and production lines; drawing process flow charts by taking raw materials, intermediate products and products as targets respectively, and summarizing and connecting the process flow charts of each reactant according to the raw material change to obtain the whole-plant process flow chart.

In other embodiments, the productivity estimation is performed after the production process flow chart is drawn, including performing device decomposition on the production process, performing device modeling based on product yield and properties, and performing the productivity estimation of device yield. Dividing a production process flow chart into a plurality of sections of curves according to different devices, modeling each section of curve according to the yield and the properties of reactants in the devices to obtain a capacity estimation model, and determining the actual capacity of each device according to the capacity estimation model.

When the method is used, all devices of the production process are decomposed and identified to obtain production devices of each step or each reactant, and capacity estimation is carried out on each production device.

The decomposition and identification comprises curve division on the production process flow chart, and the production device of each section is determined after the production process flow chart is divided into a plurality of sections according to the change of reactants.

In this case, each production device corresponds to a respective reactant, and at the same time, some reactants have the same production device. Modeling of the corresponding production plant is performed based on the product yield and properties of each reactant in the production plant.

In other embodiments, after the capacity estimation is performed, a plurality of constrained plan constraint models are established, and a plan optimization objective function is determined according to the plan constraint models, wherein the plan optimization objective function is the sum of the optimization objectives of each plan constraint model. Based on the productivity estimation model, adding constraint taking the machining index as optimization, thereby establishing a plan constraint model, and combining the plan constraint model with a plan target to obtain a plan optimization objective function.

When the method is used, after capacity estimation is carried out on each production device, a planning constraint model is established for each production device; the constraints are the corresponding constraints of the reaction process in the production device, including the material constraint and the material property constraint of each reactant, and the device yield balance, the device property balance, the device processing capacity constraint, the device fortification constraint and the public engineering balance of the reaction device.

And combining the plan constraint models of each device to obtain a plan optimization objective function, wherein the plan optimization objective function comprises sub-plan constraint models corresponding to a plurality of production processes.

In other embodiments, in the solving based on the branch-and-bound method, different constraints are respectively bound to obtain a plurality of optimal solutions, and the plurality of optimal solutions are combined to be used as an optimal planning scheme. Determining a feasible region of each constraint, and establishing a search tree containing sub-problems, namely a subset of a search space; and (3) iteratively calculating the sub-problems in a feasible domain to obtain an optimal solution of the sub-problems, updating the optimal solution of the sub-problems into a search tree, and pruning the sub-problems when the sub-problems have no solution better than the existing optimal solution.

When the method is used, a branch-and-bound method is firstly used for solving each constraint, the sub-problem at the moment is a sub-problem under each constraint, and the feasible domain at the moment is a change interval of each component under each constraint.

At this time, when the targets of different components under each constraint are used as sub-problems, if a better solution than a real-time optimal solution is generated during iterative solution, dividing the targets of the components into an exhaustive sub-problem set, then inserting the sub-problem set into branches corresponding to the components on a search tree, and carrying out the iteration of the next layer until no better solution exists, pruning the sub-problem; after all the sub-problems are explored, iteration is completed, so that a plurality of targets form branches with one component, and the branches with the plurality of components form a constrained search tree.

After obtaining the search tree of each constraint, taking each constraint as a sub-problem to form the search tree, taking all optimal solutions of each constraint as feasible domains of the constraint, taking the sub-problem at the moment as the sub-problem of the optimal solution of each constraint, then carrying out iterative solution to obtain a secondary optimal solution of each constraint, completing iteration after all the sub-problems are explored, obtaining an optimal solution combination consisting of the secondary optimal solutions of a plurality of constraints, and taking the secondary optimal solutions of the plurality of constraints as an optimal planning scheme.

In other embodiments, petroleum refining characteristic parameters are obtained, the petroleum refining characteristic parameters are cut after passing through each refining device, a plurality of main characteristic parameters and a plurality of auxiliary characteristic parameters are obtained through cutting, and an optimal solution is obtained for each main characteristic parameter and each auxiliary characteristic parameter through a branch and bound method.

When the device is used, the refining device comprises an atmospheric and vacuum device 1, a catalytic cracking device 2 and a diesel oil blending device 3 which are sequentially arranged, wherein the front input component of the atmospheric and vacuum device comprises Kewit and sand light crude oil, and the rear output component of the atmospheric and vacuum device comprises dry gas, liquefied gas, light naphtha, heavy naphtha, atmospheric residuum, normal line two and normal line three.

The naphtha and normal first line enter a catalytic cracking device 2, and the output components after the catalytic cracking device comprise wax oil cracking heavy naphtha, wax oil cracking light naphtha and wax oil cracking kerosene, wherein the wax oil cracking light naphtha enters a diesel oil blending device 3.

Meanwhile, outsourced naphtha and MTBE are added into a diesel oil blending device 3, and the output components of the diesel oil blending device comprise 0# vehicle diesel oil and 10# vehicle diesel oil.

The component entering the next device after each device is used as a main characteristic parameter, and the rest components are used as slave characteristic parameters. And solving an optimal solution by a branch-and-bound method for segmentation based on the main characteristic parameters and the auxiliary characteristic parameters.

Modifications and improvements made to the invention without departing from the spirit of the invention are within the scope of the invention as claimed.

Claims (8)

1. The production model optimization method suitable for the complex refinery is characterized by comprising the following steps of:

obtaining a general petroleum refining plan optimization model;

modeling and optimizing conversion are carried out based on the structure and characteristics of the petroleum production process;

and solving the optimal plan optimization scheme in the constraint model space by using a nonlinear programming method based on a branch-and-bound method.

2. The optimization method for the production model of the complex refinery according to claim 1, wherein the optimization transformation comprises the steps of performing variable analysis and determining the effective value range of the variable in a feasible region; performing constraint condition analysis, and screening redundant constraint and contradictory constraint according to linear correlation and nonlinear correlation among constraint conditions; and carrying out objective function analysis, and carrying out objective item balance on the contribution of the whole objective function according to each objective item.

3. The method for optimizing production models for complex refineries according to claim 2, wherein the modeling includes obtaining structural parameters and characteristic parameters of petroleum production process, wherein the structural parameters are content parameters of different component structures after petroleum is processed, and the characteristic parameters are process parameters corresponding to different processing stages of petroleum.

4. A method of optimizing a production model for a complex refinery according to claim 1, 2 or 3, wherein the basic data of the refinery process is obtained from a general petroleum refining plan optimizing model, the process index of the refinery is set as the constraint of optimization, and a production process flow chart is drawn.

5. The method of optimizing production models for complex refineries according to claim 4, wherein the step of estimating the productivity after drawing the production process flow chart comprises performing device decomposition on the production process, performing device modeling based on the product yield and the property, and performing the productivity estimation of the device yield.

6. The method of optimizing production models for complex refineries according to claim 5, wherein after said estimating the capacity, a plurality of constrained plan constraint models are established, and a plan optimization objective function is determined according to the plan constraint models, wherein the plan optimization objective function is the sum of the optimization objectives of each plan constraint model.

7. The optimization method of production model for complex refinery according to claim 4, wherein in the solving based on branch-and-bound method, different constraints are respectively bound to obtain a plurality of optimal solutions, and the plurality of optimal solutions are combined to be the optimal planning scheme.

8. The method according to claim 1 or 2, further comprising obtaining petroleum refining characteristic parameters, cutting the petroleum refining characteristic parameters after passing through each refining device to obtain a plurality of main characteristic parameters and a plurality of auxiliary characteristic parameters, and obtaining an optimal solution for each main characteristic parameter and each auxiliary characteristic parameter by a branch-and-bound method.