Several oil and gas domains raise 3D seismic imaging challenges or are only imaged with 2D seismic lines. In both cases the shape and the connectivity of faults is subject to uncertainties which may be consequential for the determination of migration paths, trap geometry and reservoir compartmentalization. Stochastic fault network simulation aims at generating a set of 3D structural models honoring prior structural concepts and conditioned by interpretations made from wells and seismic data. This set of models aims at sampling the uncertainty space related to the fault network geometry and connectivity (topology is variable from one realization to the next and emerges from the simulation process). We apply this stochastic approach to a highly-uncertain and complex fault network at reservoir scale. The used dataset is composed of several wells and 3D seismic data that poorly image the reservoir. We show how the tectonic history and the structural style can be conveyed to a stochastic fault modeling system in order to ensure the simulation of consistent 3D fault networks. We also discuss the strategy to generate suitable spatial interpretations from 3D seismic data. We then use statistical analyzes to evaluate the uncertainty about the number of faults and the number of compartments in the reservoir.