Urban Hourly Water Demand Prediction Using Human Mobility Data

Abstract

The efficient management of a water supply system requires precise water demand forecasts as inputs. This paper compares existing prediction methods and improves their performance by integrating human-related factors with water consumption in an urban area. Furthermore, a framework for processing and transforming mobility data into time-series is presented. Results show that using human mobility data improves forecasting accuracy reaching 87.6%.

Human Mobility Prediction With Region-Based Flows and Water Consumption

“…Regarding works related to water consumption data and human mobility, Smolak et al [18], [19] compares several ML and statistical models to predict the water usage based on human mobility data. The main aim of this work is the water demand forecasting of a particular area, and they actually concluded that the use of this mobility data is correlated to the water demand and this it benefits its prediction.…”

“…The human displacements are defined at a region level in an aggregated form whereas the water consumption data are defined at a much finer granularity as it is extracted directly from smart meters. Although water consumption and human-flow data have already been explored together in several prediction problems [18], [19], it is worth mentioning that their goal was the prediction of the water consumption of a population instead of its movement activity as in our proposal.…”