Sistema Iot lorawan open-source para monitoramento e gestão hídrica: implementação e validação em Campus Universitário

Abstract

This dissertation demonstrates the technical and managerial feasibility of an open end-to-end system for intelligent monitoring of water consumption on a university campus. The architecture integrates SAGATECH US-5.0 smart meters, a LoRaWAN network with a Browan gateway, the ChirpStack (v4) network server, a Python–MySQL pipeline, and Grafana visualization. The system was deployed at 182 points and operated in the operational dataset (22/09/2024–22/09/2025), producing auditable time series. Data quality was quantified along three dimensions: weighted global completeness of 78.54% (per-device median of 83.68%; Q1–Q3: 78.55–89.01%; 22.9% of points with completeness ≥ 90%); interval regularity with a median of 60.0 min and 89.9% of records within the 45–75 min corridor; and internal consistency with 0.00% negative increments and a median of zero reset events, together with P50 = 0.00L, P90 = 84.00L, and P99 = 631.00L for hourly increments. Accuracy was validated on the Dval dataset (28/04/2025–12/06/2025; 11 points) through interval-based AMI–AMM comparison without interpolation, using the timestamps closest to [t0,t1]; relative errors ranged from −0.434% to +0.459%, remaining within the ±2% limit specified by the applicable metrological regulation. As a practical deliverable, a reproducible operational triage was implemented to prioritize maintenance, based on communication silence and atypical night-time consumption (threshold defined by a high percentile of the dataset itself). The solution, in addition to being scalable and low cost, provides data quality suitable for managerial use and accuracy consistent with manual reference measurements, offering a solid basis for institutional expansion and replication.