A knowledge-and-data-driven modeling approach for simulating plant growth and the dynamics of CO 2 /O 2 concentrations in a closed system of plants and humans by integrating mechanistic and empirical models
Résumé
Modeling and the prediction of material flows (plant production, CO2/O2 concentrations, H2O) is an important but challenging task in the design and control of closed ecological life support systems (CELSS). The aim of this study was to develop a novel knowledge-and-data-driven modeling (KDDM) approach for simultaneously simulating plant production and CO2/O2 concentrations in a closed system of plants and humans by integrating mechanistic and empirical models. The KDDM approach consists of a ‘knowledge-driven (KD)’ sub-model and a ‘data-driven (DD)’ sub-model. The KD sub-model describes hourly and up to daily plant photosynthesis, respiration and assimilation partitioning using the components of GreenLab and TomSim models. The DD sub-model describes the dynamics of CO2 production and O2 consumption by the crew member using a piecewise linear model. The two sub-models were integrated with a mass balance model for CO2/O2 concentrations in a closed system. The KDDM was applied with a two-person, 30-day integrated CELSS test. This model provides accurate computation of both the dry weights of different plant compartments and CO2/O2 concentrations. The model also quantifies the underlying material flows among the crew members, plants and environment. This approach provides a computational basis for lifetime optimization of cabin design and experimental setup of CELSS (e.g., environmental control, planting schedule). With extension, this methodology can be applied to a half-closed system such as a glasshouse.