A new genetic algorithm for solving nonconvex nonlinear programming problems

Abstract

In nonlinear programming problems (especially nonconvex problems), attaining the global optimum is crucial. To reach this purpose, the current paper represents a new genetic algorithm for solving nonconvex nonlinear programming problems. The new method is simpler and more intuitive than the existing models and finds the global optimum of the problem in a reasonable time. The proposed technique, to attain the global optimum of problem (especially in large scale problems) instead of increasing the size of population which usually conduces to the curse of dimensionality – that is widespread in usual genetic algorithms – decomposes the main problem into several sub problems, but with lower size of population in each sub problem. This decomposition has been formed in such a way that it could envelop the total solution space. The proposed genetic algorithm, which determines the sufficient sub problems, could converge to global optimum of problem. To be able to measure the proposed genetic algorithm, several problems have been solved in different spectrums. Herein, it has been shown that the proposed technique, both in run time and in solution quality, is preferred to the usual genetic algorithm in this domain (nonlinear continuous programming problems).