gene expression programming algorithm

In flood conditions at the alluvial rivers with compound sections, due to momentum exchange between main channel and floodplains, flow discharge prediction by traditional methods is very erroneous. In this paper, a new method known as multi-stage gene expression programming has been used for computation of flow discharge in straight compound channels. For modeling, three dimensionless parameters of relative depth, coherence and relative calculated flow discharge, and one parameter of relative measured flow discharge were selected as inputs and output, respectively. Using 402 stage-discharge dataset from 31 laboratory and field compound channels, explicit relationships were developed for flow discharge prediction. The mean absolute errors of this method were obtained as 10.2% and 11.6%, respectively for training and testing phases. Hence compared with the Manning's formula (with mean absolute error of 19.3%), the proposed method is quit outperform. Hence, application of this method is recommended for flood flow discharge in rivers having compound channel forms. Also, by incorporating this new idea with the computation procedures of the river water surface profiles and flood routing, the design of flood alleviation schemes will be improved.