IoT for Water Management: Towards Intelligent Anomaly Detection

Given that the global water system is deteriorating and the supply and demand are very dynamic, smart ways to improve the water management system are needed in order to become more efficient and to extend the services provided to the citizens leading to smart cities. One of many water related problems that can be addressed by the Internet of Things is anomaly-detection in water consumption. The analysis of data collected by smart meters can help personalize the feedback to customers, prevent water waste and detect critical situations.

Water consumption data can be considered as a time series. Time series anomaly detection is an old topic but in this work we attempt to examine which techniques suits better for water consumption. We examine two very well-known methods for time series anomaly detection: an ARIMA-based framework anomaly-detection technique which selects as outliers those points no fitting an ARIMA process and also a technique named HOT-SAX which represents windows of data in a discrete way and then discriminates them using a heuristic.