A Multi-objective Optimization Approach Based On Differential Evolution For Operational Optimization Of Atmospheric Distillation Column

The atmospheric distillation unit is the most important separation device in oil refineries,it extracts useful component fractions(gasoline,diesel,naphtha,kerosene)from the feed by the physical reaction.The operating variables of atmospheric distillation unit(hereinafter referred to as the operating variables)determines the quality of the products and quality,the set point of operating variables effects the economic benefits of the enterprises directly.When changing the set point for each operating variable,it will often lead to increase the yield of product,and other yield of products will be reduced.Therefore the yield product is conflicted between the operating variables.We can draw the conclusion that optimization of atmospheric distillation unit is a multi-objective optimization problem with many conflicts.In addition,the distillation unit is not only a high coupling problem between the operating variables and the quality of the products,but also has the nonlinear relationship between the operating variables.It will increase the difficulty of optimization.However,multi-objective algorithms using now does not meet the requirements of atmospheric distillation unit in computational efficiency and diversity.Therefore research on multi-objective optimization approach based on for operational optimization of an atmospheric distillation unit,which meets the constraints and determines the reasonable operating variables setting so that the product yield and other performance indicators are optimized,has vital importance.Based on the problems above,supported by the National Natural Science Foundation of China "refining production process global collaborative optimization operation theory and Implementation Technology(61590922)",the research on multi-objective optimization algorithm for operational optimization of an atmospheric distillation unit has been carried out.The main work in this paper is concluded as follows:(1)Description of operational optimization and decision-making of an atmospheric tower.The operational optimization evaluation index of the distillation process(the yield of top of column,the yield of the first side and the yield of second side)are given.According to the process characteristics of atmospheric tower,the status and operating variables in the distillation process are analyzed,thus establishing the optimization constraints and decision variables,and the analysis of the focus and difficulty of the operation optimization problem was carried out.(2)An adaptive PBI decomposition technique based on differential multi-objective optimization algorithm has been proposed in this paper.In the view of multi-objective optimization framework,the algorithm is different from the previous multi-objective optimization method based on non dominated sorting.It introduced decomposition technique for solving multi-objective problems.In the stage of population update,we proposed the preferred learning strategies and then it was used to improve the convergence ability without loss of diversity.In addition,an adaptive design strategy was proposed which integrated the preferred learning strategy operator and DE/rand/1 operator with strong robustness into the process of differential evolution to enrich the information of population update.In the experimental studies,in order to verify the effectiveness of the proposed strategy,we selected two groups functions as the benchmark test problem.The results showed that the introduction of decomposition technology has improved significantly for the three objective optimization problem.The preferred learning operator and adaptive strategy have a good improvement both on convergence and distribution.The proposed algorithm has a good performance on convergence and distribution.Especially for the practical problem of solving unknown Pareto front,the adaptive strategy can gradually improve mutation strategy according to the evolution situation for adapting to different practical problem.(3)In the case of operational optimization and decision-making of distillation process,the proposed algorithm MOEA/D-SADE was applied to experimental research.Because of the strong nonlinearity and complex coupling between the operating variables and the yield of the atmospheric tower,select the relevant variables of atmospheric tower yield using maximum information coefficient method,and then established an optimization decision model with the decision variables and constraint conditions.The proposed algorithm are compared with three algorithms adopted in this papaer,the results showed that the the proposed algorithm was more effectively for operational optimization and decision-making of an atmospheric tower.Finally,through the comparison of the proposed algorithm with NSGA-II algorithm for optimization of operating variables in distillation process,we can concluded that the proposed improved PBI decomposition technique based on differential evolution adaptive algorithm can improve the operating range of the operating variable with the same yield condition.