Artificial neural network applications for optimizing water alternating gas injection: a comprehensive review

The Water Alternating Gas (WAG) process is a cyclic method of injecting alternating cycles of gas followed by water, repeating this process over multiple cycles. However, decision-making during the WAG process often involves critical choices that, at times, fall short of providing accurate and effective results. In recent years, the advent of Artificial Neural Networks (ANNs) has opened up promising opportunities to revamp Enhanced Oil Recovery (EOR) processes. This comprehensive review examines ANN-based methodologies for predicting performance of WAG injection projects. It starts with an overview of the WAG injection process, outlining its relevance in enhanced oil recovery. Subsequently, the study explores the architecture and complexity of ANNs, providing foundational insights into their functionality. The application of ANNs in developing predictive models for WAG injection is then analyzed, emphasizing their potential to enhance accuracy in forecasting WAG performance. Key aspects discussed include ANN-based WAG injection across various gas and rock types, the functions used in ANN WAG modeling, and the algorithms employed in ANN-based WAG simulations. Furthermore, the review highlights the challenges associated to implementing ANN-based predictive models in WAG projects. By synthesizing existing research, this study intends to provide valuable insights for petroleum engineers, particularly in understanding and applying ANN models for optimizing WAG injection strategies.