Predictive Modeling and Computer Vision-Based Decision Support to Optimize Resource Use in Vertical Farms

Abstract

This study evaluated several decision-support tools that can be used to create a control system capable of taking advantage of fluctuations in the price of resources and improving the energy use efficiency of growing crops in vertical farms. A mechanistic model was updated and calibrated for use in vertical farm environments. This model was also validated under changing environmental conditions with acceptable agreement with empirical observations for the scenarios considered in this study. It was also demonstrated that lettuce plants use carbon dioxide (CO2) more efficiently later in their development, producing around 22% more biomass during high CO2 conditions during the fourth-week post-transplant than in the first week. A feedback mechanism using top-projected canopy area (TPCA) was evaluated for its ability to correlate with and provide remote biomass estimations. It was shown that for a given set of constant environmental conditions, a scaling factor of 0.21 g cm−2 allowed the TPCA to serve as a rough proxy for biomass in the period prior to canopy closure. The TPCA also was able to show deviation from expected growth under changing CO2 concentrations, justifying its use as a feedback metric.