A frequent task in system design is to refine a transition system, i.e. to remove undesired transitions. This task can be hard if one is confronted with a large system. This work presents an approach for handling large systems via bisimulations: A large, possibly infinite system is transformed into a smaller system, which is subsequently refined. Finally, the refinement of the small system is transformed back into a refinement of the original large system. The treated problems include among others optimality problems and stochastic games. The last part is dedicated to an algebraic approach to bisimulations and the application of automated theorem provers.