Unconventional reservoir such as tight and shale gas reservoirs has the potential of becoming the main source of cleaner energy in the 21th century. Production from these reservoirs is mainly accomplished through engineered hydraulic fracturing to generate fracture networks that provide the gas flow pathways from the rock matrix to the production wells. While hydraulic fracturing technology has progressed considerably in the last thirty years, designing the fracturing system primarily involves judgments from a team of engineers, geoscientists and geophysicists, without taking advantage of computational tools, such as numerical optimization techniques to improve short-term and long-term reservoir production. This thesis focuses on developing novel optimization algorithms that can be used to improve the design and implementation of hydraulic fracturing in a shale gas reservoir to increase production and the net present value of unconventional assets. In particular, we consider simultaneous perturbation stochastic approximation (SPSA) and Covariance Matrix Adaptation - Evolution Strategy (CMA-ES) algorithms, which are proven very efficient in finding nearly optimal solutions. We show that with a judicious choice of control variables (continuous or discrete) we can obtain efficient algorithms for performing hydraulic fracture optimization in unconventional reservoirs. To achieve this, the hydraulic fracture production optimization problem is divided into two aspects: fracture stages placement optimization with fix stage numbers and unknown stage numbers. After check the parameters of fracture model that could be used to simulate future reservoir behavior with a higher degree of confidence, the fracture stages optimization is scheduling the fracturing sequence, and adjusting the fracture stages intensity at different locations, which is similar to well placement problem. In addition to the detailed investigation of the new optimization technique, uncertainty quantification of reservoir properties and its implications on the optimization workflow is also considered in the shale gas reservoir model. Taking into account that shale gas reservoirs are highly heterogeneous systems, stochastic optimization methods are the most suitable framework for hydraulic fracture stages placement.
