A Control Engineering and System Identification Approach to Improve a Cocoa Biofertilization Process

Agricultural production processes are complex systems with many external and internal variables involved that affect their productivity. From a systems perspective, the performance of these processes could be optimized by finding well validated mathematical models that are capable to describe their behavior. These models have been traditionally obtained via statistical methods that do not account for the time dynamic nature of agricultural processes. This paper presents a system identification and control systems approach to obtain a dynamical model for a cocoa biofertilization experiment that has been developed at the San Rafael farm in Bucay, Ecuador. The experiment is developed using a number of dosage schedules of biofertilizers for different agricultural parcels. External factors, such as ambient temperature and relative humidity, are considered in the estimation. An exhaustive search procedure is proposed to test different model orders for various parametric structures. The model is validated via percentage fits to a different set of data; a correlational residual error analysis is also presented. An improved biofertilizer dosing strategy is developed relying on the obtained model and using model predictive control (MPC) ideas. The control strategy is tested through a simulation study considering the obtained model, and under realistic conditions that resemble those from the original experiment.