Bayesian Sensitivity Analysis to Enhance History Matching Through Shale Oil Reservoirs

Many shale oil reservoirs in the US have outstanding amounts of oil reaching several billion barrels in formations such as Bakken, Eagle Ford, Niobrara, and Marcellus. These huge resources require extensive reservoir studies and development strategies, which typically conducted through dynamic simulation models. Therefore, enhancing reservoir models is essential to mimic reservoirs performance and to predict future behavior under different constrains. In this research, Bayesian sensitivity analysis was considered to optimize the process of reservoir model history matching and to obtain the most representative reservoir model. Oil production from shale reservoirs requires the use of horizontal wells with hydraulic fractures and this technology affects reservoir performance significantly. As a result, more parameters will be enforced in the reservoir simulation models leading to exacerbate the efforts of achieving history matching. In this research, Bayesian Model selection (BMS) was applied as a stochastic solution to flag the key parameters that affect the reservoir performance and history matching dramatically. The results obtained from the best model showed that horizontal matrix permeability, fracture half-length, and rock compressibility are more sensitive than others. The novelty of Bayesian Model Selection comes from its procedure to optimize the best fitting model among 50 different models based on Bayes' theorem.