This paper presents a study of the application of a neural network to forecast storm runoff volumes from a catchment, the Upper Bukit Timah (UBT) catchment, in Singapore. A widely used catchment model, SWMM , was applied to generate 3 sets of 273 runoff volume data from the UBT catchment. Each set of the 273 data was simulated for one storm event at various values of the calibration parameters. This data set was required in the catchment calibration method suggested by Ibrahim and Liong to construct a response surface relating an objective function and the calibration parameters. A back-propagation neural forecaster with a non-linear Sigmoid activation function was trained to mimic the SWMM simulations. After training, the forecaster was tested on SWMM's simulation results of seven other storms. The comparisons show that the trained neural forecaster not only can predict the runoff volumes in a much shorter time but also results in low prediction errors.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £34 +P&P)