Natural fractures are commonly observed in the unconventional reservoir. Production history indicates that natural fractures have been playing an important role in the oil and gas development progress by improving the permeability of the reservoir and increasing the well productivity. In addition, inappropriate development strategies result in the unreasonable single well oil rate, early water breakthrough, severe damages to the unconventional reservoir and overwhelming economic losses when the fracture properties and distributions are not well understood before the development. Hence, it is of great importance to propose a powerful and efficient workflow to describe the fracture distribution clearly, including building a 3D fracture model, performing history matching and forecasting productions of the unconventional reservoir. In this study, we present a powerful and practical workflow through using Fracflow software and EDFM (Embedded Discrete Fracture Model) to build the 3D DFN (Discrete Fracture Network) model. The main methodology used to perform the fracture modelling allows rigorously handling of both hydraulic fractures and natural fractures that can be identified in an unconventional reservoir. This modelling allows computing the real geometrical fracture attributes (mainly orientation and density) and the spatial distribution of fractures. Fracture conductivity values will be calibrated through a comparison of the Kh(permeability thickness) from the well test to the Kh model computed from the upscaling of the fracture model. The mentioned model above will be built by means of a stochastic simulation constrained by the results of the static and dynamic fracture characterization. In the reservoir simulation phase, EDFM processor combining commercial reservoir simulators is fully integrated to perform history matching and production performance forecast of the unconventional reservoir. With a new set of formulations used in EDFM, the non-neighboring connections (NNCs) in the EDFM are converted into regular connections in traditional reservoir simulators, and the NNCs factors are linked with gridblock permeabilities. EDFM provides three kinds of NNC pairs, transmissibility factors, and the connections between fractures and wells. With the aid of the EDFM processor, we can obtain the number of additional grids, the properties of fracture grids, and the NNCs as the simulation input. From the proposed workflow, complex dynamic behaviors of natural fractures can be captured. This will further ensure the accuracy of DFMs and the efficiency offered by structured gridding. The practical workflow for the unconventional reservoir from modelling to simulation highlights the model constrained by the results of the static and dynamic fracture characterization, and the high efficiency to model discrete fractures through the revolutionary EDFM processor. Through this workflow, we can perform history matching effectively and simulate complex fractures including hydraulic fractures and naturally fractures. It potentially can be integrated into existing workflow for unconventional reservoirs for sensitivity analysis and production forecasting.