FertilizeSmart: Exploiting IoT and Differential Evolution for Optimizing Crop Fertilization

Efficient crop fertilization (e.g., nitrogen) is crucial for maximizing agricultural productivity, ensuring food security, and promoting sustainable farming practices. Traditional methods, such as fixed-rate fertilizer applications or soil sampling, often result in inefficiencies, over-fertilization, and environmental harm, as they fail to account for dynamic in-season weather conditions and varying nutrient needs at different crop growth stages. In this work, we introduce FertilizeSmart, an innovative framework that optimizes crop fertilization by leveraging Internet of Things (IoT) technologies. The goal is to determine the optimal fertilization strategy throughout the season. To this purpose, at the core of FertilizeSmart, is an optimization problem that maximizes crop yield while minimizing the amount of fertilizer used. The crop yield in response to different timings and rates of applied fertilizer is estimated using a process-based crop simulation model, namely the Decision Support System for Agrotechnology Transfer (DSSAT). The optimization problem is then solved periodically, by an improved Differential Evolution (DE) algorithm that trades off exploration and exploitation of available solutions, throughout the crop growth cycle, following a Model Predictive Control (MPC) approach. This adaptive approach allows FertilizeSmart to respond to dynamic weather conditions and adjust fertilizer application to meet varying nutrient demands across growth stages. Moreover, we perform extensive simulation experiments and results show that FertilizeSmart significantly outperforms existing fertilizer recommendations, achieving yields approximately 20% higher while reducing fertilizer usage by up to 32% compared to the fixed application rate.