A hierarchical control system for optimizing crop production in a vertical farming container and its Internet of Things architecture

This work presents a hierarchical control system to maximize profit for vertical farming in a shipping container. In its upper optimization layer, a model predictive control (MPC) approach was implemented to calculate the optimal crop cycle duration, the optimal duration of the artificial lighting periods, and the optimal air temperature setpoints, as variables with high impact on crop growth and directly linked to resource consumption and costs. In addition, to address commercial contract aspects, a minimum weight constraint to be met by the crop at the end of the cycle was incorporated. The MPC strategy uses a temperature model and a crop growth model, both based on first principles, which were calibrated and validated using climate data and destructive samples of lettuce crops from a real container located in southeastern Spain. For practical deployment, the hierarchical control system was conceptually implemented in a cloud platform under the Internet of Things (IoT) paradigm, which provides interoperability with sensors and actuators, remote supervision, and scalability to multiple containers. The responses provided by the hierarchical system are analyzed after three simulations involving the achievement of the optimal setpoints in a lower control layer, a sudden change in the minimum crop weight restriction, and a case in which the optimal setpoints are not reached during the crop cycle. In each simulation, the performance of the MPC strategy is compared against a baseline controller that operates with manually fixed setpoints. The proposed control system achieves production targets five days earlier, reducing energy consumption by 6.5% and increasing profitability by 2%. It would allow for an additional crop growth cycle each year, increasing container productivity (and therefore annual profits). Moreover, it is able to comply with variability in product demand, while the baseline controller would violate the maximum cycle duration constraint and might not yield any benefits. Regarding performance with uncertainty, the proposed control system can generate adequate setpoints to continue optimizing production, which shows that the system automates decision-making to assist farmers in achieving long-term objectives. • Treats photoperiod duration as a decision variable, enabling cost minimization. • Ensures contract compliance under demand shifts, unlike fixed-setpoint controllers. • Cuts crop cycle duration by five days, saves 6.5% energy, and boosts profit by 2%. • Allows for an additional crop cycle each year, increasing annual productivity. • Automates decisions using first-principles models validated with real lettuce crop.