| On The Convergence Of An Evolutionary Algorithm, Particle Swarm Optimization (PSO) And Its Application Besiana Cobani1; Aurora Ferrja1; 1UNIVERSITY OF TIRANA, Tirana, Albania; PAPER: 451/Mathematics/Regular (Oral) SCHEDULED: 12:35/Sat. 26 Oct. 2019/Hermes (64/Mezz. F) ABSTRACT: The evolutionary methods are optimization methods that converge to the global solution. There are many optimization techniques nowadays used and the one we are working is the evolutionary method PSO. Many authors have proposed various modifications of the basic PSO parameters with the goal to obtain a variant of PSO with best performance algorithm complexity. In our case, first, we present a modified PSO algorithm. Then we analyze the convergence of the proposed algorithm using differential equations. More precisely we relate a difference equation with a differential equation, and study the behavior of its solution. The solution brings results for the parameters of PSO, specifically for the coefficients of acceleration. Since the PSO results depend on its parameters, we propose new parametersthem based on the convergence study. We give an application in the energetic field in Albanian case, in the main three hydropower cascades of the country, which consist of three hydro power plants. |
