This paper presents a versatile methodology to calculate parameters that characterise the demand of a DMA. This parameters are used for the leak modelling so that a predictive model is built and trained with historical data and used to detect on-line new leaks so that the repair time can be reduced applying proper leak localisation techniques. This methodology has been programmed using R where the modelling packages available provide assortment of predictive models easily to implement. It has been integrated in a Data analysis tool in order to utilise the great amount of information coming continuously from the WDN. Once the methodology and the tool are described the results applied to a real DMA are presented. This work has been carried out by the fundació CTM Centre Tecnològic (CTM) collaborating with Research Center for Supervision, Safety and Automatic Control (CS2AC) within a research project of Aigües de Manresa.

Peer Reviewed