Case Study Artificial Intelligence (AI) Assisted Numerical Reservoir Simulations in a Mature Reservoir Under Waterflood Development in the Sultanate of Oman

PDO is transforming its field development planning by adopting digital technologies and Artificial Intelligence (AI) to improve organizational efficiency and maximize business value through swift quality decisions and an evergreen forecast. In this context, the company has approached a number of third parties to bring in solutions in this domain. In 2021 one such collaboration with 3rd party contractor to test a novel solution involving data driven AI based dynamic simulator in a mature brown fiend setting. The objective was to test the tool, the efficacy and efficiency of the process, robustness and ease of use and its utility in current setting [1]. Existing dynamic modelling workflows with conventional simulators are extremely time consuming to update and upgrade in a mature brownfield setting. These conventional and lengthy iterative process of working might leave value on the table. It is time consuming to update history match with all the extra inputs and forecast; and optimizing the development with all the input parameters within a short timeframe is always a challenge. The process employed in this approach was based on deep learning artificial neural networks (ANN) coupled with numerical simulators and along other static model inputs. The reservoir static and the flow dynamics were used as feature parameters to train the ANN, while the historical field production was used as the target parameters. The ANN training exercise identified the most contributed static and dynamic parameters to the historical production; therefore, these main parameters were given a higher weight in production forecasting and reservoir management. This AI-simulation method was expected to be faster, data driven and allow a faster testing of multiple development strategies in short time. This paper outlines the experience of an AI-assisted numerical simulation approach to unlock the potential of brown oil fields in south Oman by reducing the time spent on modelling and base case anchoring. It also enables evergreen forecasting by integrating AI techniques with numerical simulation. The AI-simulation was tried in a brown field with an existing FDP generated using conventional simulation tool where >50% of the FDP propose wells have been drilled. The outcomes from the AI-simulation result were compared with conventional simulation and with Actual field performance. Optimization was also conducted to locate the sweet spots for future drilling and WRFM opportunities. This optimized workflow has the potential to enable step change improvements in time and value for brownfield development and optimization for future developments.