Author(s)

A. Candelieri & F. Archetti

Abstract

Big Data opportunities arise from high rate data streams acquired through smart sensors and smart meters, which, even for small water utilities, may produce a huge amount of data to be stored. This data enables the application of new data analytics to infer reliable predictive functionalities, with implications ranging from reducing No Revenue Water (NRW) to optimizing the water-energy nexus, meeting ever more pressing budgetary constraints. This paper presents the approach proposed in the EU-FP7-ICT project ICeWater, combining time series clustering, for the identification of typical daily urban water demand patterns, and Support Vector Regression for performing a short term forecast. Promising results obtained on the Water Distribution Network (WDN) in Milan are presented. The approach has been designed to also be applied on smart metering data related to individual customers, addressing Big Data analytics issues. Keywords: smart water management, predictive analytics, short-term demand forecasting.

Keywords

smart water management, predictive analytics, short-term demand forecasting.