Sistema IOT de baixo consumo para detecção de vazamentos de água baseado em LoRaWAN e FIWARE: desenvolvimento e avaliação experimental

Abstract

In the context of intelligent water resource management, Internet of Things (IoT) systems have shown great potential for monitoring water consumption and detecting leaks. However, achieving high energy autonomy in field devices remains a challenge, especially for low-cost and largescale solutions. This work proposes a complete, scalable, and energy-efficient IoT architecture for smart water metering and automatic residential leak detection, integrating hardware, firmware, communication, and cloud-based processing. The prototype was designed with a strong focus on energy efficiency, using the STM32L031F6P6 microcontroller and the SX1262 transceiver, operating mainly in low-power modes and powered by a LiPo battery with energy harvesting support. Water meter pulses are collected by the IoT node and transmitted via the LoRaWAN protocol. Cloud communication is established through The Things Network (TTN) and integrated into the FIWARE infrastructure. The system is hosted in an AWS environment, where the analytical models are continuously and automatically executed, receiving data directly from the FIWARE database and providing real-time inferences through an integrated API. In the intelligent layer, unsupervised machine learning and spectral analysis techniques are applied to identify anomalies and microleaks. The Elliptic Envelope model is used for detecting anomalous leak events in time series data, while the Short-Time Fourier Transform (STFT) enables the recognition of cyclic low-flow patterns associated with microleaks. Experimental results demonstrated high accuracy in anomaly detection, with an average F1-score of 0.88 and average power consumption below 1 mW, resulting in prototype autonomy exceeding 130 days without recharging.