The Research For Forecasting Urban Water Demand

As the population of city increases and the economy develops rapidly, water demand becomes more and more large. The existing water resource and facilities can't meet the future demand. The contradiction between supply and demand is becoming more and more remarkable. In order to change the present serious situation, finding new wellspring and saving water must be made to instruct the far future water supplying infrastructure construction. However, all these programs are based on city water demand forecast. Thus, water demand forecast carries an important sense towards the development of water facilities in city.The urban water demand forecasting can be divided into four parts: industrial water demand,agriculture water demand,life water demand and municipal water demand. Structure analysis about water demand is conducted, based on researching the history and economy of a city. And related factors for water demand are studied deeply. Under the help of SPSS software, correlation analysis of factors is made and six factors are chose as the critical factors of city water demand forecast.BP network (namely, Back-Propagation network) is a kind of artificial neural networks. It can approach any consecutive function, just designed with one hidden layer. This nonlinear character makes it gain a wide application room in forecasting field. However, the traditional BP algorithms holds many drawbacks, such as: being prone to fall into a local minimum,training with a slow speed,being poor in generalizing,being hard to decide the number of hidden layer units. This paper makes an improvement for BP on the basis of the analysis of defect root.Matlab software is an excellent develop environment which supports system's building from conception design,building model,simulation to realization with real time. It is also characterized as many merits, such as: programming rapidly, powerful drawing performance, being easy to visualize,express operation (esp. when operating with an matrix or an array ). In addition, powerful BP network toolbox and express interface building made an advanced impulse for it towards information system develop.Basing on all considerations above, this paper builds an urban water demand forecasting model with BP network,offers its solution and conducts an prediction for a schedule year. The experimental results showed that this model owned a high precise value towards prediction and can meet prediction's demand in urban water planning.Finally, an urban water demand forecasting system is developed on a platform of Matlab. The system, which holds many excellent practical values, also offers an simple and effective way for software develop and possesses some good instructions towards other optimizing software in water supply and drainage system.