Differential Evolution with Tabu List for Global Optimization and Its Application to Phase Equilibrium and Parameter Estimation Problems

Abstract

Stochastic global optimization and their applications are attracting greater attention and interest in the recent past as they provide better solutions with relatively less computational effort. Among the many popular methods, differential evolution (DE), proposed by Storn and Price [J. Global Optim. 1997, 11, 341−359], is a population-based direct search algorithm for nonlinear and nondifferentiable functions, and has found numerous applications due its simplicity, ease of use, and faster convergence. In this work, we attempted to improve the computational efficiency of DE further by implementing the concept (i.e., avoiding revisits during the search) of tabu search (TS) using the tabu list in the generation step of DE; it also provides diversity among the members of the population. DE with tabu list (DETL) is initially tested on several benchmark problems involving a few to thousands of local minima and 2−20 variables. It is then tested on challenging phase equilibrium calculations followed by parameter estimation problems in dynamic systems known to have multiple minima. The results show that the performance of DETL is better compared to DE and TS for benchmark, phase equilibrium, and parameter estimation problems.