The research is constructed by the storm sewer mode of University of Illinois, using back-propagation neural network to achieve the optimization design. The thesis uses the observation values, from May to October in 2007, gained by the Central Weather Bureau and applies neural network to estimation models to train them. By tracing their tracks, we can find out a better outcome. It is always an important task to do the research for designing the estimated cost before the sewer construction starts, especially in the system of the pumping stations and network of the sewers. The result of the research also shows that applying backpropagation neural network to the system and simplifying the mathematical modes increasingly can greatly improve the time of computerizing process and its accuracy.