Robust sample-based model predictive control of a greenhouse system with parametric uncertainty

Achieving optimal resource use efficiency is a key challenge in modern greenhouse production systems. Optimal performance in terms of crop yield and resource efficiency can in theory be achieved via optimal control. Standard optimal controllers are not designed to deal with uncertainty, whereas considerable model prediction errors occur due to the mismatch between the model and the real system. This paper explores the relation between parametric uncertainty, and performance with respect to crop yield, CO2 demand, ventilation demand, and heating energy. This is done using the following steps 1) extension of an existing controller model with parametric uncertainty, 2) design of a sample-based robust model predictive controller and 3) analysis of control performance under increasing parametric uncertainty. The results predict that control performance is significantly sensitive to parametric uncertainty. A relative parameter uncertainty of 20%, reduced crop yield with 11% compared to the case without uncertainty. Furthermore, a 20% uncertainty decreased CO2 demand with 80%, whereas it increased ventilation demand with 96%, and increased heating energy demand with 90%.