Applications of robust multi-objective genetic algorithm (RMOGA) for robust optimization of chemical processes in the petroleum industry.

摘要

Oil refinery decision makers are usually encountered with a large number of business and engineering decisions. These decisions are usually not integrated rather they focus only on their individual aspects. As a result of this, most of oil refineries undergoes either with business decisions or with engineering decisions. This is particularly challenging when strategic decisions towards plant profitability and sustainability have to be made under uncertain environment. Also, the refinery decision support systems (DSSs) are not developed enough to integrate all decision-making processes of refinery. To facilitate the efficient decision-making process for dashboard decision support system, robust multi-objective genetic algorithm (RMOGA) with online approximation under interval uncertainty is a useful tool when used in a decision support framework. The traditional approach looks at all the inputs as deterministic and therefore the optimum solutions can be sensitive to uncertainties. The goal of the RMOGA is to obtain optimum solutions which are relatively insensitive to variation in objective and constraint function due to uncertainties present in input variables/parameters.;In this work we demonstrate the applications of the developed approaches in RMOGA decision support system, i.e. nested RMOGA and sequential RMOGA in petroleum industry to a number of refinery and gas processes such as Natural Gas Liquids (NGL), Liquefied Natural Gas (LNG) and Steam Methane Reforming (SMR) units. For this study we modeled the engineering aspects of the processes using the HYSYS simulator. The overall objectives are to find the best values of minimum energy and maximum recovery under uncertain environment. The sequential AA-RMOGA was performed on the all the case studies and the results showed the 9.87% energy reduction for NGL fractionation train and best value of ethane recovery for selected constraints set. The same approach for the SMR process showed 25.4% energy reduction with best value of hydrogen recovery under uncertain parameters and constraints set. The compression section and pre-cooling section of LNG plant is selected for the application of sequential AA-RMOGA. The results obtained shows the energy reduction of 12.8% from the baseline plant but it was 4% less than the reduction obtained with the simple genetic algorithm in some work. If we consider LNG liquefaction section and other section along with the compression and pre-cooling we can get more reduction in energy with many other operational alternatives.;This decision support framework provides robust optimum solutions that are insensitive to the presence of uncertainties for the operational parameters of processes discussed in this work. It provides the decision makers with an effective tool that utilizes their expertise in decision-making. The solution obtained shows that suggested technique is good enough than the traditional approaches for optimizing integrated network under interval uncertainty.