Research On Genetic Algorithm Applying To Drought-Flood Predicting

The drought-flood disasters have done great damage to social ecnomy in Hebei area. It is vital for us to do research in the field of drought-flood predicting, for grasping its inherent law. The genetic algorithm(GA) is a very promising artificial intelligent technique to solve some problems in this field,such as identifying model coefficients.After comparing classical binary GA and real GA,the author presents a new version about storing chromosome ,or binary chromosome machine-level stored,and takes the first step of developing two basic bit genetic operators compiled by C language, basic crossing-over bit operator and basic mutation bit operator with strong ability to embed.This paper discusses four approaches of time series analysis. Those are classical time series analysis, markov model,grey predicting, artificial neural network (ANN). In terms of indemnifying model coefficients ,the author develops two GA versions embedded the two bit genetic operators, for evaluating classical time series models and the weights of ANN.The two versions are applied in modeling on the base of Shijiazhang rainfall datas in flood period of many years, and are compared with traditional methods. The results show that inproved GA is very efficient ,and GA-ANN is a robust,flexible way among drought-flood predicting approaches. What's more.this paper does more research in the generalizing ability of ANN.