Comparison of Particle Swarm Optimization, Genetic, and Dragonfly Algorithm to Invert Vertical Electrical Sounding

Abstract

Summary In direct current (DC) sounding (VES) data, inversion technique is applied to obtain the electrical property (resistivity) of subsurface. Inversion of VES data is a non-linear problem. Therefore, non-linear inversion techniques are used in this research, i.e. Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Dragonfly Algorithm (DA). PSO and GA are frequently used by researchers to solve VES inversion problem. However, DA has never been applied for that problem. These algorithms combine the explorative (global search) and exploitative (local search) concepts to obtain global optimum solution. Initially these algorithms are tested for noise free and added synthetic data in order to assess their ability. In both synthetic data inversion result, these algorithms are able to obtain model parameters which are sufficiently close to true model. In convergence rate, PSO and GA tend to be exploitative (fast convergence rate), while DA tends to be explorative (low convergence rate). In field data, these algorithms are applied to identify the sediment layer in the dried lake area, around lake Ayamaru, Sorong, Papua. In inversion result, the sediment layer has the thickness and resistivity value repectively about 3.3 to 3.5 m of dried lake surface and 20–27 Ωm.