  Citation: Silva, M.G.; Madeira, S.C.; Henriques R. Water Consumption Pattern Analysis Using Biclustering: When, Why and How. Water 2022, 14, 1954. https://doi.org/10.3390/ w14121954 Academic Editor: Pilar Montesinos Received: 12 May 2022 Accepted: 16 June 2022 Published: 18 June 2022 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affil- iations. Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). water Article Water Consumption Pattern Analysis Using Biclustering: When, Why and How Miguel G. Silva 1,2, * , Sara C. Madeira 1 and Rui Henriques 2 1 LASIGE and Faculdade de Ciências, Universidade de Lisboa, 1749-016 Lisboa, Portugal; lasige@ciencias.ulisboa.pt 2 INESC-ID and Instituto Superior Técnico, Universidade de Lisboa, 1000-029 Lisboa, Portugal; info@inesc-id.pt * Correspondence: mmgsilva@ciencias.ulisboa.pt; Tel.: +351-217500532 Abstract: Sensors deployed within water distribution systems collect consumption data that enable the application of data analysis techniques to extract essential information. Time series clustering has been traditionally applied for modeling end-user water consumption profiles to aid water manage- ment. However, its effectiveness is limited by the diversity and local nature of consumption patterns. In addition, existing techniques cannot adequately handle changes in household composition, dis- ruptive events (e.g., vacations), and consumption dynamics at different time scales. In this context, biclustering approaches provide a natural alternative to detect groups of end-users with coherent consumption profiles during local time periods while addressing the aforementioned limitations. This work discusses when, why and how to apply biclustering techniques for water consumption data analysis, and further proposes a methodology to this end. To the best of our knowledge, this is the first work introducing biclustering to water consumption data analysis. Results on data from a real-world water distribution system—Quinta do Lago, Portugal—confirm the potentialities of the proposed approach for pattern discovery with guarantees of statistical significance and robustness that entities can rely on for strategic planning. Keywords: water consumption analysis; biclustering; time series; pattern discovery; clustering; subspace clustering; water distribution systems 1. Introduction Sustainable management of water supplies generally depends on the continuous collection, monitoring, and analysis of sensor data (e.g., pressure, flow, consumption), which need to be translated into usable information for daily control and strategic planning. Over the last few years, with the arrival and deployment of smart grid meters within water distribution systems (WDSs), there has been an increasing collection of data that raises new opportunities and challenges for the entities responsible for managing these systems [1]. The data produced by smart meters, usually in the form of georeferenced time series data (measurements sequentially recorded through time), provide essential information that enables the application of data analytics’ tools to model end-use water consumption profiles. With this actionable information, water companies and municipalities have better knowledge of what to expect from customers and thus develop efficient marketing strategies [2], promote water-saving behavioral changes [3], enhance water infrastructure planning [4], and manage water demand and detect anomalies [5]. In the literature, a considerable number of clustering approaches have been proposed for the analysis of water consumption time series. Laspidou et al. [6] applied cluster- ing on water-billing data to distinguish household from business end-use consumers; Cheifetz et al. [7] proposed an enhanced clustering methodology to discover consump- tion profiles from time series data; Ioannou et al. [8] also presented a technique to de- tect behavioral patterns in water consumption, grouping users by behavioral similarities; Water 2022, 14, 1954. https://doi.org/10.3390/w14121954 https://www.mdpi.com/journal/water