Multi-objective Optimization Approach For Cooperative Operation Indices Of Prefractionator And Atmospheric Distillation Column Based On Decomposition Method

Crude oil distillation unit is the first process in crude oil processing.Crude oil is divided into petroleum products such as gasoline,kerosene,diesel oil and processed raw materials for follow-up units by physical distillation.The tower is the core equipment of the whole process of crude oil distillation unit.The performance and operation state of the tower affect the product quality and yield of the atmospheric and vacuum unit.At present,the most commonly used crude oil distillation unit processes in most refineries are three-column processes,which are:prefractionator,atmospheric distillation column and vacuum distillation column.The three columns are coupled and interact with each other.The atmospheric pressure tower is a product separation device,the operation of the prefractionator is closely related to the quality and yield of the side line of the atmospheric distillation column.Improper operation of the prefractionator will lead to the failure of the non-production of oil or black oil in the side line of the atmospheric distillation column,therefore,the cooperative modeling method of the prefractionator and the atmospheric column is adopted.The output of the side line of atmospheric distillation column is closely related to the economic index of the refining enterprise,and there is a conflict relationship between it,which is a multi-objective optimization problem.There are many constraints in oil refining,the operational variables and targets are highly coupled and nonlinear,and the operational efficiency of the crude oil distillation unit has a great impact on the subsequent units.The traditional gradient optimization methods are related to the initial value and requires functional behavior,the existing multi-objective evolutionary algorithms are slow in convergence and large in computation,so it does not meet the requirements of the optimization problem of oil refining operation index.Therefore,it is of great significance to propose an effective multi-objective evolutionary algorithm for solving highly nonlinear complex problems.In view of the above problems,this paper,relying on the project of National Natural Science Foundation of China,"The Theory and realization Technology of Global Cooperative Optimization of Refining production process(61590922)",has carried out the multi-objective optimization research on the cooperative optimization of prefractionator and crude oil distillation unit.The main work of the study is as follows:(1)The synergistic process of prefractionator and atmospheric column was analyzed and described.This paper introduces the research status and operation optimization algorithms of the existing crude oil distillation unit,describes the specific technological process and the role of each part of the prefractionator and the atmospheric pressure column,and expounds the coordination problem between the prefractionator and the atmospheric pressure column.In this paper,the conflict relation between the side line output is analyzed,the operation variables and the final objective function of the model selection are given,and the difficulty of the operation optimization of the crude oil distillation unit and the optimization problem in this paper are discussed.(2)An improved multiobjective optimization algorithm based on decomposition method is proposed.In view of the large feasible range and target solution space of operational variables in refining process,the diversification of oil products and industrial fuel demand,and the diversification of operation modes of product production schemes,the operation speed of crude oil distillation unit variables affects the operation of subsequent units.According to the characteristics of the generation strategy of MOEA/D itself,the sub-problem is uneven and the local search ability is poor,a multi-operator and multi-parameter strategy is proposed to generate the offspring.The improved decomposition of the comparison Update part with uniform comparison strategy and the neighbor individual renewal part with local search strategy The standard optimization algorithm(MOEA/D-ML)improves the diversity and convergence of the algorithm,and improves the robustness and convergence speed of the algorithm.In order to test the performance of the improved algorithm,it is applied to the DTLZ,ZDT,WFG,MOEADDE F and UF five groups of 38 standard test functions.On this basis,the saliency analysis,parameter sensitivity analysis and effectiveness analysis of three strategies are carried out.The experimental results show that the proposed MOEA/D-ML algorithm has a fast convergence rate in the evolution of population,and can converge quickly when the number of function evaluation is less.The solution set of uniform distribution can be obtained at the approximate real Pareto front,and the algorithm has good distribution and convergence.(3)Combined with the operation optimization problem in crude oil distillation unit,the factors related to the output of side line products are qualitatively analyzed from the point of view of process mechanism.According to the specific process of the plant and the experience knowledge of the technicians in the workshop,thirteen variables of great influence are selected as the input variables of the model,under the constraints of the operation index of the plant and the quality.The yield of the side line product,which measures the processing technology of the index,is selected as the output of the model.Based on the actual data of the plant,the side line yield and constraint model of the prefractionator and the atmospheric distillation column are established,and several algorithms are combined with the feasibility criterion to deal with the constraints.These optimization algorithms with constraints are applied to the practical side line yield model of prefractionator and atmospheric distillation column.Because it is impossible to know the true optimal solution of the cooperative operation optimization model,the solution set distribution plot and the hypervolume evaluation index are used to study the performance of several optimization algorithms applied in practice is evaluated.The results obtained from the four optimization algorithms verify that the improved decomposition multi-objective optimization algorithm proposed in this thesis has good optimization results.