The exploitation of gas from tight gas reservoirs has been increasing due to the advances in unconventional gas technologies and depletion of conventional gas resources. Drilling horizontal wells and placing transverse hydraulic fractures are needed to produce economically viable amounts of gas from tight sands. In this research, optimization of the design of hydraulically fractured horizontal well placed in naturally fractured tight gas sand reservoir systems was studied. A commercial reservoir simulator is coupled with artificial neural network (ANN) to create an expert system that can be used to design an efficient stimulation strategy. Reservoir simulation was used to generate production profiles of the reservoir system and, then, used to train the artificial neural network system. The ANN tools developed in this project consists of two parts: forward and inverse processes. In the forward process, an input data set that includes well and completion design parameters and reservoir properties is used to create an ANN toolbox to predict the production profile. In the inverse process, several design parameters and the desired production profiles were introduced into the input data to create an ANN toolbox to estimate the optimum design parameters. It is found that in both expert systems, functional links play a significant role in the success of the ANN tool. The results estimated by this developed toolbox can be used as preliminary information about the expected production profiles or the required wellbore and fracture design parameters for different cases.
This study shows that natural fracture permeability is the principal factor that determines the magnitude of the production. In addition, overall conductivities of the natural fractures and the hydraulic fractures are important factors affecting production.