The improvement and application of real-coded multiple-population genetic algorithm

This paper first examines the theory of Real-Coded Genetic Algorithm(RCGA) and put forward that locus independent selection strategy should be adopted when performing RCGA crossover operation. According to this strategy a new crossover operator is designed and tested. Secondly, the paper pays great emphasis on the scalability of the GA algorithm when selecting GA strategies, thus the Multiple Population Genetic Algorithm (MPGA) is selected as the framework of the algorithm and two population-level strategies are designed and tested. Combining the new crossover operator and these two strategies together, the paper at last comes to the improved MPGA and it is then applied to the multi-layer gravity inversion of both experimental models and field data of Xinjiang area,China.