The Research Of Genetic Algorithm Assimilation System In The MM5 Model

In this paper, a genetic algorithm is combined with the mesoscale NWP model MM5. We take a heavy rainfall process during 14-15, June, 2002 as the experimental research object, and carry out data assimilation experiment of actual observation data. To conquer the problems of "lower efficiency in searching" and "premature convergence" in canonical genetic algorithms, traditional Genetic Algorithm is improved , new GA operators and fitness functions and algorithm with hybrid operator, which use the self-adaptive operator to improve the optimization performance and hybrid operator to increase diversity degree of the population, are designed. The results show that in the improved assimilation experiments of actual observation data. whichever assimilation system are used, the initial field of numerical prediction models are improved and the prediction effect of physical field and rainfall are somewhat enhanced. Because the model computational quantity is very huge, so in the experiment the genetic algorithm is paralleled, the result suggests that the computational speed of program is largety heightened, based on the upper conclusion the genetic algorithm assimilation system will be widely applied in the complex model.The genetic algorithm is new assimilation method, we believe that along with the theory progress of atmosphere science and mathematics, and the computational capability, increase of computer, the better prediction result will be gained with the use of four-dimension data assimilation system of genetic algorithm.