Abstract This paper describes the experience gained from the development and application of data-driven modeling techniques to enhance the efficiency of waterflood operations in several mature fields in Block 61, Ecuador. A modeling application that integrates data analytics, machine learning (ML) and physics-based reservoir modeling has been prototyped since 2022 as part of a wider digital transformation initiative. The goal is to significantly enhance decision-making efficiency, reduce analysis time, and support proactive field management strategies. The application is designed to help multidisciplinary teams by simplifying complex reservoir behavior and fostering accessibility for both experts and non-experts in secondary recovery. After over 50 years of production by pressure depletion, these partially exploited, logistically challenging, oilfields have been actively developed since 2017 using horizontal well technology and waterflooding. Initially water injection allocation was decided using simple analytical methods with grossly simplifying assumptions and spreadsheet calculations that were found to be ineffective and tedious as complexities in the management of the waterfloods began to unfold. After exploring various alternatives and experimental prototyping, including pure AI/ML, numerical reservoir simulation, surrogate modeling, management decided to develop a hybrid AI-Physics application based on capacitance-resistance and fractional flow theories to replace the spreadsheet. The new smart injection allocation system supports seamless data ingestion from a corporate database that stores daily reconciled production and injection data. An interactive web-based interface empowers users to evaluate multiple optimization scenarios, perform what-ifs with short-term forecasting capability to support decision-making. Leveraging several complementary digital technologies, the system has been instrumental in offering new insights, increasing efficiency and recovery through informed decision-making. Analysis time has reduced from several hours to a few hours, but more importantly, blind testing has proven the application to be remarkably dependable. Since its deployment in September 2023, the application has contributed to an estimated incremental cumulative oil production of over 128 KBO across the fields analyzed, with an economic impact of 3.81 MMUSD to date. In addition to operational gains, the system has improved cross-team collaboration, enabling faster, more consistent responses to field conditions. Its deployment also contributes to sustainability goals by minimizing CO₂ emission through decreased need for on-site monitoring.
Mejía-Arango et al. (Mon,) studied this question.