Model-Based Design and Evaluation of a Multi-Variable Hybrid Predictive Control Loop for Precision Fertigation Using Multi-Sensor Data Fusion

Abstract

Efficient water and nutrient management remains a major challenge in modern agriculture due to rising fertilizer costs, global water scarcity, and environmental concerns such as groundwater contamination from nutrient leaching. Conventional fertigation typically relies on uniform application rates that neglect spatial and temporal root-zone variability, resulting in inefficient resource utilization. To address these limitations, this study proposes and evaluates an automated Multi-Variable Hybrid Predictive Control (HPC) framework for precision fertigation. The system captures the highly nonlinear dynamics of rapid soil moisture changes and slower nutrient transport using a multi-variable state-space model that combines continuous physical processes with discrete control logic, including pump scheduling and fertilizer injection. The predictive controller is enhanced by an Extended Kalman Filter (EKF)-based Multi-Sensor Data Fusion (MSDF) framework, which mitigates measurement noise and sensor drift from capacitive moisture and electrical conductivity (EC) sensors through adaptive covariance estimation, providing reliable root-zone state estimates. A 30-day closed-loop MATLAB simulation demonstrates robust tracking performance and disturbance rejection. Following the receding horizon strategy, the controller suspended irrigation and fertilization during rainfall events, effectively utilizing natural precipitation and eliminating the risk of groundwater nutrient leaching during precipitation events. Compared with conventional uniform fertigation, the proposed approach achieved a 25% reduction in water use and a 30% reduction in fertilizer consumption. Overall, the proposed framework provides an effective solution for improving resource-use efficiency while supporting environmentally sustainable precision agriculture.