The Application of Artificial Intelligence Methods in Water-Flood Reservoirs: Current Status, Technical Comparison, and Prospects

With the gradual deepening of oilfield development, water injection optimization has entered a new stage of real-time and intelligentization. The artificial intelligence technologies have gradually become the core means of waterflood reservoir research. This paper systematically investigates four key issues in waterflood development, including automated history matching, injection/production fluidity analysis, identification of water flow dominant channels and injection/production optimization. Then, the current applications of artificial intelligence methods in these fields are analyzed. Next, from the technical point of view, the intelligent technologies used in waterflood development are summarized into three categories: image-to-image, im-age-to-sequence and sequence-to-sequence modeling framework. This paper systematically compares and analyzes the strengths and limitations of these modeling methods. Furthermore, the comprehensive performance of various techniques is evaluated according to the indicators of accuracy, computational efficiency, robustness and interpretability. Finally, this paper discusses the development direction of intelligent waterflood optimization. In particular, the potential for applications of knowledge and data fusion-driven technologies is emphasized, which can promote the intelligent application of water-driven development oil field. Therefore, more efficient and accurate reservoir management and real-time intelligent waterflood optimization strategies will be ultimately realized.