:Authors

Najeh Alali, Mahmoud Reza Pishvaie1, Vahid Taghikhani

Abstract

 The production of highly viscose tar sand bitumens using Steam Assisted Gravity Drainage (SAGD) with a pair of horizontal wells has advantages over conventional steam flooding. This paper explores the use of artificial neural networks (ANNs) as a valid alternative to the traditional SAGD simulation approach. Feed forward, multi-layered neural network meta-models were trained through the back-error- ropagation (BEP) learning  algorithm to provide a versatile SAGD forecasting and analysis framework. The constructed neural network architectures were capable of satisfactorily estimating the recovery factors of the SAGD production as an enhanced oil recovery method. Rigorous studies around the hybrid static-dynamic structure of the proposed network were conducted to avoid the overfitting phenomena. The FANN-based simulations were able to fairly capture the underlying relationship between several parameters/operational conditions and rate of bitumen production, which proves that ANNs are a viable tool for SAGD simulation 

Keywords

Artificial Neural Network, Meta-modeling, Enhanced Oil Recovery, Steam Assisted Gravity Drainage