A short-term forecasting method for District Metered Area (DMA) water consumption that first identifies other DMAs whose consumption is correlated (by Pearson correlation coefficient) with a target DMA, then uses those correlated patterns together with, or in place of, the target's own historical data as input to an LSTM deep learning model. A real-world study on five DMAs in Limassol, Cyprus shows the deep model beats a classical statistical baseline, that a DMA can be forecast from its correlated DMAs alone (enabling forecasting for unmonitored or sensor-faulty DMAs), and that adding correlated-DMA data improves accuracy even when the target's local data is available.
Accurate water consumption forecasting helps utilities ensure reliable supply, optimise operations, and plan infrastructure. Urban water distribution networks are divided into District Metered Areas …