Avaliação do desempenho do algoritmo de evolução diferencial na solução de problemas de otimização dinâmica

Abstract

The solution of problems of dynamic optimization (POD) involves determining the profile of one or more of a control variable which minimize or maximize a given performance index. Except for a few problems when analytical methods are applicable, the most dynamic optimization problems is nonlinear and requires a numerical solution approach for obtaining the profile of the control variable. Classical methods of numerical solution based on the principle of differential and integral calculus use information derived from Objective Function and constraints to determine an optimal solution. Heuristic natural optimization methods have in common the random nature of the search for an optimal solution. These are methods that do not use information derived from Objective Function and restrictions. The Differential Evolution algorithm developed by Storn and Price (1995) is of simple implementation and only requires the definition of some parameters. Among some of the features that the algorithm presents Differential Evolution, is the ability to manipulate functions targeted multimodal nonlinear and non-differentiable, and is a method that has a high probability of finding the global optimum. In this work we present solutions for different dynamic optimization problems. The use of Differential Evolution algorithm simplifies the POD solution avoiding the need, for example, for the solution of problems at the boundary value or significant algebraic manipulations in solving special problems. The results obtained by Differential Evolution are compared with results obtained by other authors using traditional methods and the natural methods. The analysis of the results demonstrate the efficiency of the application of the Differential Evolution algorithm in solution of different dynamic optimization problems.