This study focuses on improving the productivity of a fed-batch ethanol fermentation process by developing and implementing in real time, an optimum feeding policy. It is well-known that fermentation processes are typically very complex, involving different mechanisms of transport processes, cellular metabolism, and complex biochemical reaction networks. A constraint-based stoichiometric model is developed using the systems biology approach, which can quantitatively predict the cellular behavior of the E coli strain, KO11. However, the predictions from this model are accurate under low concentrations of substrate. In order to extend the usage of the model to higher substrate concentrations, we modified this model by introducing changes in glucose utilization rate to restrict the metabolic capacity of the cell. Using the proposed model, the fed-batch optimization problem becomes a constrained optimization problem, and a modified Iterative Dynamic Programming (IDP) algorithm was developed with an adaptive-stage updating methodology, which was applied to solve the global optimization problem. Applying the proposed optimization method, a feed profile was generated. The objective function maximized the ethanol productivity (g/l-h) and minimized the acetic acid production. Experiments were carried out based on the optimal feed profile generated using the modified stoichiometric model and the Iterative Dynamic Programming algorithm. Approximately 90% of the theoretical ethanol yield was obtained; the fermentation consumed a total of 521.5 g of glucose and produced 237.5 g of ethanol.