The Improved Genetic Algorithm and the Application of It in MM5 Model of Four-Dimensional Variational Data Assimilation System

This article applies an improved genetic algorithm into a MM5 model of four-dimensional variational data assimilation problem. It provides a new and rather effective method to initialize the field of numerical prediction. On the basis of analysis on variational assimilation theory and simple genetic algorithm, we have ameliorated selection operator combining the characteristic of variational problem on their own. Besides, Elite Technology is adopted to speed up the convergence rate at the same time. Meanwhile, we can also improve mutation operator and use dynamic mutation probability. We increase the mutation probability, break down the stagnation and accelerate the optimization process when the optimization process bogs down due to the lack of diversity. Finally, we check up the MM5 model of four-dimensional variational data assimilation system ground on improved genetic algorithm. Compared with the simple genetic algorithm, the experiment indicates that the system has a stronger assimilation ability to raise the accuracy of MM5 forecasting.