Offshore development tends to aim for high production with fewer wells, which leads to large well spacing and wide range of low-pressure gradient. Compared with onshore, the low-velocity nonlinear flow, matrix stress sensitivity, and CO2 miscibility have significant impacts on the injection optimization design. However, existing studies rarely examine the impact of these characteristics on injection parameters, including injection fluid selection, injection rate and timing. Therefore, we first establish a 3D embedded discrete fracture model for compositional simulation. Multiple mechanisms, including low-velocity nonlinear flow, matrix and fracture stress sensitivity are considered comprehensively. After evaluating the sweep area, injection capacity, and production performance of different injection fluid (H2O, CO2, N2), a fluid selection strategy was proposed. The cumulative oil production under a certain water cut/gas-oil ratio is taken as the objective function. Then the injection rate and timing are studied. Results show that, when the injection rate of CO2 and H2O is high, the fluid breakthrough comes early. When the injection rate is low, the reservoir pressure is poorly maintained, resulting strong matrix stress sensitivity, and low CO2 miscibility efficiency. Therefore, there is an optimal CO2 and H2O injection rate. In comparison with the traditional Darcy flow, the optimal H2O injection rate is higher. The maximum difference reaches 11.1% with matrix permeability of 5 × 10−3 μm2. As the matrix permeability increases, optimal CO2 injection timing is gradually advanced. While the optimal H2O injection timing is gradually delayed. The proposed 3D-EDFM compositional simulator can serve as an effective and reliable method for injection-production optimization.