Plant metabolism is characterized by a wide diversity of metabolites, with systems far more complicated than those of microorganisms.　Therefore, plant metabolic engineering is still a challenge compared to microbial engineering.　Metabolic engineering requires system-level understanding of metabolism and mathematical modeling is useful for understanding dynamic behaviors of plant metabolic systems.　Time-series metabolome data has great potential for estimating kinetic model parameters to construct a genome-wide metabolic network model.　However, data obtained by current metabolomics techniques does not meet the requirement for constructing accurate models.　In this article, we highlight novel strategies and algorithms to handle the underlying difficulties and construct dynamic in vivo models for large-scale plant metabolic systems.　The coarse but efficient modeling enables predictive metabolic engineering and also the prediction of unknown mechanisms regulating plant metabolism.