Using the transformation method

the restraint of traditional projection model can be removed. The unconstrained projection model is optimized by artificial fish swarm algorithm and used to calculate the projection vector values of flood. The obtained results are comparable with the actual case. The Shijiao Station of the Beijiang River is taken as an example in this paper. The results indicate that this new model is more effective and reasonable than the traditional method for flood classification. During the period from 1954 to 2006

the flood in 2006 was the biggest one

the floods in 1964

1966

1968

1982

1994 and 1997 also showed higher intensity. This suggests that the flood situation should not be despised in the Biejiang River. The flood control work should be strengthened in order to ensure the flood control ability of the dike in the Beijiang River.