Short-term water demand forecasting using artificial neural networks: IIT Kanpur experience

In this paper the relatively new technique of artificial neural networks (ANNs) has been investigated for use in forecasting short-term water demand. Other methods investigated for comparison purposes include regression and time series analyses. The data employed in this study consist of the weekly water demand at the Indian Institute of technology (IIT) Kanpur campus, and the rainfall and maximum temperature from the City of Kanpur, India. The ANN models consistently outperformed the regression and time series models developed in this study. An average error in forecasting of 3.28% was achieved from the best ANN model. It was found that the water demand at IIT Kanpur is better correlated with the rainfall occurrence rather than the amount of rainfall.