Big Data Analysis of Residential Water Demand Records

PhD candidate Robert Enouy and Associate Professor Andre Unger from the Department of Earth and Environmental Sciences will present, "Big Data Analysis of Residential Water Demand Records," on Thursday, November 23 at 2:30 p.m.

Light refreshments will be provided.

Please register.

More information

The City of Waterloo, Ontario, Canada applies a volume-constant pricing structure and services upwards of 30,000 residential accounts during bimonthly billing periods. Consumption data for 60 periods between January/February 2007 and November/December 2016 were collected with the total dataset comprising 1,549,371 observations that correspond to 51,291,348 of billed water for residential consumers. We develop a general methodology to reproduce the water demand histograms from each billing period as parametric probability density functions (PDFs) that are asymmetric, shifted, and tail-weighted. The versatility of the methodology is illustrated by application of a consistent PDF parametrization to the entire set of residential water demand histograms. The parametrization is able to reproduce the location, scale and shape of the histogram data as PDFs with a compression efficiency in excess of 98%. We infer a causal relationship between parameters that control the shape of the residential water demand PDF to price and weather, with weather being comprised of precipitation and temperature components. Moreover, the PDF representing the entirety of the residential water demand from all active accounts can be viewed as a continuum changing shape in response to these price and weather signals. The causal relationships are derived using simply the first term of a Taylor series expansion of each parameter of the PDF on price and weather, which is equivalent to linear regression. To improve the water demand forecast, we include additional terms in the Taylor series expansion to capture effects such as the increase in non-consumptive water use during summer months that occur despite price increases. We demonstrate the inclusion of these cross-correlation terms significantly increases the accuracy of the mean water demand forecast to extreme seasonal weather conditions.

Speaker bio

Robert Enouy is a PhD candidate at the University of Waterloo, Ontario, Canada, Earth and Environmental Science Department. He received his BASc and MSc degrees in chemical engineering from the University of Waterloo, specializing in subsurface multiphase/ multicomponent flow in unconsolidated porous media. Robert has six years of environmental consulting experience, specializing in groundwater remediation, human-health risk assessment, and contaminated plume analysis. Enouy has a keen interest in using a multidisciplinary approach to better understand highly complex problems in water science.

Andre Unger is an associate professor in the Department of Earth and Environmental Sciences at the University of Waterloo. His research interests span: financially sustainable water infrastructure, pricing risk in engineering projects; computational finance for pricing derivatives and bonds, and multi-phase compositional simulation of hydrogeological and petroleum reservoir applications. He received his BASc in geological engineering as well as MSc and PhD degrees in Earth Sciences all from the University of Waterloo.