Study on Genetic Algorithm of Changeable Structure and Its Application to the Lag Time of Watershed

In order to overcome the limitations of binary-encoded genetic algorithm, an improved real-encoded genetic algorithm of changeable structure was proposed, and then the algorithm was applied to establish the model of the lag time of watershed. In the model, the rainfall intensity, rainfall amount,antecedent precipitation index, and location of rainfall centroid were adopted as input indexes of watershed rainfall, and the lag time of watershed was adopted as output index. Dapoling, Changtaiguan and Xixian subwatershed of Huaihe river basin were selected as studied area, the model results show that the relation between the lag time of watershed and properties of watershed and rainfall in three nested subwatersheds are identical, and the mean absolute error of the lag time of watershed is less than 1.5h for the established model; the proposed real-encoded genetic algorithm of changeable structure is superior to the binary-encoded in precision.