A Case Study of Digital Twin for Greenhouse Horticulture Production Flow

Abstract

Greenhouse horticulture production is associated with high uncertainty and a long learning process due to its high dependency on the outdoor & indoor environment and plant types. Digital Twin (DT) technology enables a faster understanding of greenhouse horticulture facilities, obtaining insight into the production process flow and investigating the consequences of production decisions. However, no digital twin has been developed in this field due to the complexity of greenhouse production. Therefore, this paper presents a case study of a DT development for a Danish greenhouse production flow using multi-method modeling and multi-agent simulation. The results show that the developed DT can accurately represent the greenhouse production process and estimate the plant growth state with an absolute error of 0.31 days compared to the observed production. Furthermore, the developed DT can accurately predict deviations to the plant growth state corresponding to previously observed behavior at the facility. To capture the greenhouse production process flow at the top-level greenhouse DT agent, the underlying physical agents developed included: compartments, growth climate, conveyors, staff, tables, plants, soil machine, table loading, and packing station as well as the packing station. Lastly, the developed DT method supports agent re-usability for other case studies.