Energy Optimization in Hydroponic IoT Systems Through Deep Sleep and Aggregate Water Stress Index Control

Agriculture increasingly depends on Internet of Things (IoT) architectures able to reduce energy consumption while preserving reliable irrigation decisions and transparent operator interaction. This paper proposes and validates an adaptive IoT architecture for hydroponic lettuce cultivation, implemented on the M5Stack Tab5 platform with an ENV Pro BME688 sensor, a substrate-state probe, a miniature irrigation pump, local Aggregate Water Stress Index (AWSI) algorithm computation, and telemetry persistence in Microsoft Azure Cosmos DB through Azure Functions. The contributions consist of introducing a correlation mechanism between monitored microclimatic parameters, the applied irrigation regime, the novelty of the proposed infrastructure, the introduction of a unique AWSI algorithm, and the new mechanism for energy efficiency. The proposed AWSI algorithm generates adaptive decisions for activating the irrigation pump, optimizing water consumption, and the energy profile of the system. In the experimental dataset, 1994 records were analyzed. A revised Azure telemetry validation export included 697 records collected between 27 May and 25 June 2026; after excluding 30 incomplete start-up or sensor-zero entries, 667 complete records were used for controller-consistency analysis. The energy contribution is quantified through battery autonomy tests. Continuous operation depleted the 14.8 Wh Tab5 battery after 10 h 30 min, corresponding to an equivalent average power of 1.41 W. The proposed deep sleep strategy extended autonomy to 25 h and reduced equivalent average power to 0.59 W, which represents a 138.1% autonomy increase and an approximately 58% reduction in average power demand. The AWSI-to-irrigation response also confirmed progressive actuation. These results indicate that the proposed architecture contributes to energy-efficient hydroponic control by combining local autonomy, cloud traceability, adaptive actuation, and a human-centric embedded dashboard.