Optimizing Flow Monitoring with AI: A Cost-Effective Approach for Municipalities

Flow monitoring is an essential tool for assessing the capacity of collection systems. However, its high cost often strains the budgets of modeling and planning projects, forcing municipalities to reduce data collection efforts. This reduction can take the form of shorter monitoring durations—potentially missing critical storm events—or fewer monitoring locations, both of which can compromise the quality of hydrologic assessments. This study presents an AI-driven approach that significantly reduces flow monitoring expenses—by up to 70%—while maintaining the collection of long-term flow data and ensuring high spatial resolution.

The proposed approach leverages machine learning to reconstruct flow data at temporary monitoring locations for periods when the meter was absent, using long-term data from strategically placed permanent flow meters. First, total monitored flow is segregated into baseflow and storm flow components. One AI model is trained to regenerate the baseflow, including the diurnal pattern of the temporary meter, using the baseflow data from the permanent meter. A second AI model utilizes spatially distributed rainfall data and storm flow from the permanent meter to estimate storm flow at the temporary location. By summing the baseflow and storm flow components generated by these two models, the total flow at the temporary meter is estimated.

The accuracy of the AI model is validated using multiple goodness-of-fit measures to ensure high prediction accuracy across varying storm configurations. The goodness-of-fit measures include the normalized Nash Stucliffe and the normalized Kling-Gupta efficiency measures. Once trained, the AI models are applied to generate total flow data for the temporary meter during periods when it was not in operation.

This approach enables municipalities to optimize their monitoring strategies by deploying permanent meters at fewer strategic locations while rotating temporary meters across multiple sites for short-term data collection. The results demonstrate that AI-driven flow reconstruction is a cost-effective alternative to long-term monitoring, providing a scalable solution for improved stormwater and wastewater management.