Comparison of Multivariate Regression and Artificial Neural Networks for Peak Urban Water-Demand Forecasting: Evaluation of Different ANN Learning Algorithms

For the past several years, Cyprus has been facing an unprecedented water crisis. Four options that have been considered to help resolve the problem of drought in Cyprus include imposing effective water use restrictions, implementing water-demand reduction programs, optimizing water supply systems, and developing sustainable alternative water source strategies. An important aspect of these initiatives is the accurate forecasting of short-term water demands, and in particular, peak water demands. This study compared multiple linear regression and three types of multilayer perceptron artificial neural networks each of which used a different type of learning algorithm as methods for peak weekly water-demand forecast modeling. The analysis was performed on 6 years of peak weekly water-demand data and meteorological variables maximum weekly temperature and total weekly rainfall for two different regions Athalassa and Public Garden in the city of Nicosia, Cyprus. 20 multiple linear regression models, 20 Levenberg-Marquardt artificial neural network ANN models, 20 resilient back-propagation ANN models, and 20 conjugate gradient Powell-Beale ANN models were developed, and their relative performance was compared. For both the Athalassa and Public Garden regions in Nicosia, the LevenbergMarquardt ANN method was found to provide a more accurate prediction of peak weekly water demand than the other two types of ANNs and multiple linear regression. It was also found that the peak weekly water demand in Nicosia is better correlated with the rainfall occurrence rather than the amount of rainfall itself. DOI: 10.1061/ ASCE HE.1943-5584.0000245 CE Database subject headings: Artificial intelligence; Forecasting; Municipal water; Regression analysis; Water demand; Water resources; Neural networks. Author keywords: Artificial intelligence; Forecasting; Municipal water; Regression analysis; Water demand; Water resource management; Artificial neural networks.