Solving the Constrained Nonlinear Optimization Based on Greedy Evolution Algorithm.
Junhong Si,Kaiyan Chen,Sen Zhang,Yipeng Guo,Bao Zhang
DOI: https://doi.org/10.1109/icnc.2011.6022167
2011-01-01
Abstract:In order to improve the local convergence of differential evolution algorithm, we puts forward the greedy evolution (GE) algorithm based on the greedy search strategy. According to the fitness value and the selection probability, the population of a generation is classed best vectors, better vectors and poor vectors. The best vectors is retained in the child population, the better vectors is replaced if the newly generated vector in its neighborhood is better than objective vector, and the poor vectors is regenerated until the new vector is not worse than the objective vector. Improving the locally search ability and ensuring the diversity of the population, the convergence of GE increases obviously. Analysis of 3 test problems, the reasonable range of controlling parameters is determined: NPS is 1-2 times than NP, δ is 0.05-0.3, and SP is 0.4-0.8. Comparing the optimum solution of GE algorithm with differential evolution and particle swarm optimization, the result shows that GE is better than others.