Data-Driven Proxy Models for Improving Advanced Well Completion Design under Uncertainty

In order to improve the design of advanced wells, performance such wells needs be carefully assessed by taking reservoir uncertainties into account. This research aimed develop data-driven proxy models for simulation and assessment oil recovery through under uncertainty. An artificial neural network (ANN) was employed create accurate computationally efficient as an alternative physics-based integrated well–reservoir created Eclipse® simulator. The speed accuracy compared physic-driven were then evaluated. evaluation showed that while developed are 350 times faster, they can predict production unwanted fluids with a mean error less than 1% 4%, respectively. As result, considered tool uncertainty analysis where several simulations need performed cover all possible scenarios. this study, applied quantification from completed different types downhole flow control devices (FCDs). According obtained results, other well completion design, autonomous inflow valve (AICV) technology have best in limiting able reduce associated risk 91%.