From hourly readings to daily leak flags: an approach to post-meter leakage detection and analysis in residential water consumption

This work presents a data-efficient procedure for distinguishing between leak and no-leak days using standard hourly smart meter readings. The approach relies on a fixed Minimum Night Flow window and classifies days based on the presence or absence of continuous registered volume during night hours. Applied to more than 21,000 users over one year, it is validated through the application of a Generalised Linear Mixed Model, which confirms that leak days exhibit systematically higher hourly volumes across the usage groups. The procedure also quantifies the hourly and daily impact of leakage and characterises its temporal organisation and influence on diurnal consumption patterns. Using only routinely available hourly data, it provides utilities with a practical framework for assessing post-meter leakage and enhancing the operational value of existing smart-meter infrastructures.