Streamline simulation is simplified proxy for more complex, full-field simulation approaches. One useful information derived from streamlines is the relative connections between injectors and producers. When we search for improved target rates on which to set injectors and producers, we use this network to approximate local derivatives, avoiding the highly inefficient approximation in traditional derivative-free optimization methods. We compare the results to FloodOPT, a heuristic-based approach that makes use of the Injector/producer pair information to improve oil production, as well as global pattern search, the formal but computational expensive method.