The effectiveness of surfactants to improve oil recovery in unconventional oil reservoirs through wettability alteration and enhancement of spontaneous imbibition has been extensively studied. However, very few simulation studies have considered the impact of complex fracture networks on the wettability alteration due to the difficulty in handling the connection between fractures and matrix. Non-uniform fracture networks are difficult to represent using traditional discrete fracture models and local grid refinement (LGR). In this study, we used the Embedded Discrete Fracture Model (EDFM) method to simulate wettability alteration in reservoirs with complex fracture networks. In EDFM, the reservoir is discretized with structured grids, and additional grids are created to represent the fractures. Each fracture plane is that is physically embedded inside a matrix grid is discretized by the cell boundaries, and transport coefficients are calculated for those additional grids. We verified our EDFM model against LGR in both oil wet and water wet conditions for relatively uniform geometries, observing good matches in well productivities and pressure responses. Then we performed a series of numerical sensitivity studies to demonstrate the impact of natural fracture networks on the wettability alteration. Our results show that fracture density, sets of fractures, and fracture conductivity significantly affect the well productivity. The matrix permeability also has similar effects. We modified the wettability condition of the activated fracture grids and matrix grids around fractures from oil-wet to water-wet. The impacts of fractures and permeability were greatly amplified in changing from oil-wet to water-wet systems, as occurs when surfactant is added to injected frac water. This study improves our capability in the area of modeling complex fracture networks, and provides meaningful guidance for enhanced oil recovery in shale reservoirs when designing a surfactant injection strategy.