A modified genetic algorithm with sampling technique for distillation sequence synthesis

Abstract

A number of chemical engineering optimization problems generally involve highly nonlinear function, which contains numerous local optima in the feasible area. These problems are very difficult to solve and to obtain the global solution. Genetic algorithm (GA) is an optimization method widely used to solve complex optimization problems because GA can successfully solve these difficulties. Moreover, GA is easy to implement. However, GA has some shortcomings that are premature convergence and weak exploitation capabilities. The major reason of its drawbacks causes initial population lacking uniformity properties. In this work, we develop a new efficient genetic-based optimization algorithm by introducing sampling techniques to select a good set of initial population. These are Latin hypercube sampling (LHS), Faure sequence sampling (FSS), and Hammersley sequence sampling (HSS). The performance of the proposed algorithms and a simple genetic algorithm (SGA) is compared in terms of solution quality and speed of convergence to the global optimum through several complex optimization problems and a case study. The case study involves distillation sequence synthesis of methanol/water system. The objective of this problem is to find the suitable sequence and operating points providing a maximum profit. From the results, with the same parameters, our proposed techniques provide a better solution than SGA and/or converge to the global solution more than four times as fast as SGA.