Central Prediction System for Time Series Comparison and Analysis of Water Usage Data

• Published in: IEEE access 2020, Vol.8, p.10342-10351
• Main Authors: Ji, Mingeun, Yi, Gangman, Jung, Jaehee
• Format: Article
• Language: English
• Subjects: additive regression model; Analytical models; ARIMA; Autoregressive models; Autoregressive processes; Data models; Drinking water; Error detection; Flow velocity; Leak detection; Moisture content; Non-revenue water; Pipes; Predictive models; Regression analysis; Regression models; Revenue; Rural areas; Statistical analysis; time series; Time series analysis; Urban areas; Water consumption; Water flow; water leakage alert system; Water pipelines; Water resources; Water supply
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2963373

Revenue water flow is defined as the amount of water for which the water rate has been collected, against tap water production, whereas non-revenue water (NRW) is defined as water that has been produced, but for which payment cannot be charged. In South Korea, there are big differences in NRW among the regions, and the NRW ratio in urban areas is higher than in rural regions. To reduce regional differences and effectively manage the water system, a management system to lower the NRW ratio is required. In particular, the NRW ratio can be reduced through an automatic leakage detection and sensor-error automatic checking system for feed water pipes and piping in household, and through leakage detection of water supply and drainage pipes that transport large volumes of water. Therefore, this study develops a system that can generate automatic alarms whenever abnormal usage is predicted via analysis of household water flow rate. Linear regression, ARIMA model, and additive regression model are compared to find the best method with high accuracy. The proposed method can support efficient water system management to lower the NRW ratio.