Water demand estimation and outlier detection from smart meter data using classification and Big Data methods

Abstract

Automatic Meter Reading (AMR) systems are being deployed in many cities to obtain insight into the status and the behavior of District Metering Area (DMA) with more granularity. Until now, the water consumption readings of the population were taken one per month or one each two-months. In contrast, AMR systems provide hourly readings for households and more frequent readings for big consumers. On the one hand, this paper aims at predicting water demand and detect suspicious behaviors – e.g. a leak, a smart meter break down or even a fraud – by extracting water consumption patterns. On the other hand, the main contribution of this paper, a software framework, based on Big Data techniques, is presented to tackle the barriers of traditional data storage and data analysis since the volume of AMR data collected by Water Utilities is enormous and it is continuously growing because this technology is expanding .