Systematic analysis for the redirection of carbon flux in metabolite-producing microorganisms requires the comprehensive understanding of their complex metabolic processes. The use of large-scale dynamic models of metabolism plays key role in the understanding of these processes and the study of possible metabolic engineering interventions. However, the generation of such models is hampered by the intrinsic nonlinearities of enzymatic reactions, and the uncertainties at different levels. In particular there is limited knowledge about the exact kinetic mechanisms, and many of the parameters involved in these mechanisms remain largely unknown. In this study we propose a systematic methodology to generate large populations of dynamic non-linear models of metabolism using the ORACLE (Optimization and Risk Analysis of Complex Living Entities) framework. Instead of seeking for an optimal value of the estimated kinetic model parameter values, we integrate thermodynamics, available omics, and kinetic data to construct populations of models that are locally stable, and consistent with the observed physiology. To demonstrate the utility of this methodology we constructed a population of large-scale dynamical models of optimally grown E. coli. We used these models to perform large-scale perturbations of E. coli metabolism and to analyze their simulated responses. The aforementioned analyses provides valuable insight for the design of metabolic engineering strategies towards amplification of desired product-forming pathways.