Greenhouse Gas Emissions Optimization for Vegetable Processing in Production Area Using Deep Deterministic Policy Gradient

Abstract

The continuous expansion of cold chain logistics has drawn increasing attention to its associated greenhouse gas (GHG) emissions. To support the cold chain's low-carbon transition, this study focuses on GHG reduction in handling processes within vegetable production area. We first employed sensitivity analysis to identify key parameters for GHG reduction across different handling stages. Subsequently, GHG emissions optimization model was developed using the Deep Deterministic Policy Gradient. Handling processes in vegetable production were modeled as a Markov decision process. Through reward function design, the model achieves multi-objective optimization by simultaneously minimizing GHG emissions, maximizing vegetable mass retention, and shortening on-farm storage time. A case study demonstrates that processing 1000 kg of vegetables under the current system generates 117.06 kg carbon dioxide equivalent (CO2eq), whereas the proposed optimization model reduces the emissions to 36.63 kg CO2eq while simultaneously achieving a significant increase in the vegetable mass retention rate.