Data-Driven Modeling of Switched Systems and Model Predictive Control for Irrigation Management

This paper proposes a data-driven method for modeling and optimizing irrigation processes. It employs a specialized auto-regressive model that considers delays and switching behaviors. In the control architecture, discrete inputs are connected to each irrigation unit, which improves their management individually. The proposed Model Predictive Control (MPC) manages multiple irrigation units, which complies with the actuator operational constraints. This results in a non linear mixed-variable optimization problem. To address this, linearization methods are applied. Then, the Gurobi optimizer is used for efficient computation. The results of the simulation using field data show that this approach leads to optimal irrigation schedules.