Data-Driven Assessment of Water Meter Errors for Replacement Planning

Accurate water metering is essential for reducing non-revenue water (NRW) and supporting sustainable urban water management. This study proposes a data-driven framework for assessing water meter performance and supporting replacement planning by integrating advanced metering infrastructure (AMI) data with semi-supervised learning. Using case studies from two Chinese cities, an apparent metering error (AME) metric was developed by combining flow-dependent error curves with actual usage patterns. A neural network model, trained on 136 mechanical meters and applied to 140,000 residential meters, showed that aging meters systematically under-register water use (AME: −6% to −2%), mainly because of low-flow measurement failures and prolonged service duration. Results indicate that optimized replacement strategies can reduce greenhouse gas emissions and operating costs compared with conventional practices. The proposed framework provides a scalable approach for utilities to improve apparent loss assessment and meter replacement planning