With a high demand of oil, optimization of oil output in the reservoir field is urgent and significant. Practical production problems often involve large, highly complex reservoir models, with up to thousands of unknowns and many constraints. Moreover, our understanding of the reservoir is always highly uncertain, and this uncertainty can be reflected in the reservoir models. Consequentially, performance prediction and performance optimization, which are the ultimate goals of the entire modeling and simulation process, depend a lot on the data assimilation process and uncertainty from reservoir models themselves. In this thesis, the main goal is to discuss two ways of optimizing oil production and do research on the uncertainty of optimization according to data assimilation and reservoir models. Chapter 2 states an efficient method of data assimilation—Asynchronous Ensemble Kalman Fliter and its application in a square reservoir model. Chapter 4 and 5 discuss the two aspects of production optimization respectively: optimization of control of well settings and optimization of well placement. I use gradient-based method and implement it into simsim simulator, which is a simple simulator of reservoir model. In Chapter 5, I apply Particle Swarm Optimization method to optimize well placement given a fixed well settings. In Chapter 6, I do a research on the uncertainty of optimization results due to data assimilation and then study on the optimization process given a true permeability and porosity field. I found a stable performance of the optimization when applying field data estimated by data assimilation and demonstrate the effects of two types of optimization. Some disadvantage and future recommendations are presented in the last part.