Smart solar-powered hydroponic system based on artificial intelligence as a replicable model for sustainable food development in vulnerable communities

Elizabeth Puelles Bulnes M.; Delgadillo R.; Aguilar-Alonso I.; Luis Inche Mitma J.; Espinoza V.A.A.; Atoche Puelles A.L. · Proceedings of the LACCEI international Multi-conference for Engineering, Education and Technology (2026) · DOI: 10.18687/laccei2026.1.1.2757

Resumen

This study presents the design, implementation, and pilot experimental evaluation of a compact Nutrient Film Technique (NFT) hydroponic system integrated with photovoltaic solar energy, IoT-based monitoring, and intelligent control, intended for domestic and community applications as well as environments with limited water availability, space, or energy infrastructure. The objective was to preliminarily assess the technical feasibility of a modular and potentially replicable model capable of improving water-use efficiency, energy management, and crop operation through automation and continuous monitoring. As a distinguishing feature, the system incorporates an Intelligent Microgreenhouse Health Indicator (IISM), complemented by a Random Forest machine-learning model for classifying the operational status of the crop and supporting user monitoring and decision-making. A prototype was developed using a Raspberry Pi, environmental and physicochemical sensors, actuators, and an autonomous photovoltaic energy system. The pilot evaluation was conducted over 30 days after transferring six lettuce (Lactuca sativa) seedlings, with an initial height of approximately 6-8 cm, into the Microgreenhouse. During this single pilot cycle (n = 6), the system achieved an average fresh weight of 280 ± 10 g per plant and a total production of 1.68 kg per cycle, representing a descriptive increase of 16.7% compared with a previous manual reference of 1.44 kg per cycle. Water consumption was 15 L per cycle, compared with 54 L for the manual reference, corresponding to a 72.2% reduction. The energy subsystem operated fully off-grid and achieved 16 h of autonomy without photovoltaic input. The Random Forest model achieved an overall accuracy of 91.2%, with a Precision of 0.9047, Recall of 0.8996, and F1-score of 0.9018 for operational-state classification. During intelligent system operation, manual intervention was also reduced by 55%. These results should be interpreted as preliminary due to the limited sample size, the use of a single experimental cycle, and the fact that the manual comparison corresponded to a non-concurrent historical reference. It is concluded that the integration of NFT hydroponics, solar energy, IoT, and artificial intelligence represents a technically viable alternative for developing compact and sustainable food-production systems for domestic and community settings, particularly in contexts with constraints related to space, water availability, or energy infrastructure.

Instituciones: Universidad Nacional Mayor de San Marcos; Universidad Nacional del Callao