Fluid flow simulation is used to optimize oil and gas production. Naturally or hydraulically fractured reservoirs hold a significant part of reserves, difficult to assess. Fractures may create preferential flow paths heavily impacting fluid flow. Accurate modeling of fractured media accounting for strong contrasts would allow operators to optimize resources exploitation while better controlling environmental risks. Integrating sparse available data, we aim at predicting fluid flow processes occurring in the earth’s subsurface accounting for multi-scale fractures with a simply parameterized model. Improving the computational time and results reliability, we propose a full integrated strategy suitable for fractured reservoir specificities by simulating the fractures complexity on large scales. The techniques developed in this thesis, whose interest is demonstrated in an unconventional field case study, can find other applications in geothermal engineering and water resources management