Data-driven based investigation of pressure dynamics in underground hydrocarbon reservoirs

The process of storing natural gas in geological formations involves applying pressure to force the gas into and out of the porous and permeable reservoir. In response to gas extraction/withdrawal and storage/injection, the reservoir compresses and expands as a major consequences of fluid pore pressure variations. The major challenge associated with this type of energy systems is learning the pore pressure variations within the grid as fluid is being injected and/or withdrawn. As such, it is essential to identify a realistic model that accounts for the pore pressure variations at any point in time. In this paper, we present a data-driven technique called Dynamic Mode Decomposition (DMD) to investigate the pressure dynamics of an underground hydrocarbon reservoir model in relation to natural gas injection/storage. For demonstration purpose, we first implement a hydrocarbon reservoir model using a benchmark data of the first Society of Petroleum Engineers (SPE1) Comparative Solution Project. Applying DMD to a pressure field data of the simulated reservoir model, we show that DMD is capable of approximating the average reservoir pressure change and pore pressure variations within the reservoir grid over time with up to 99% accuracy. Given that depleted reservoirs are already developed hydrocarbon reservoirs, we conclude that DMD could serve as a reliable tool for fast evaluation of pressure dynamics of underground natural gas storage given its low complexity and insignificant loss of prediction accuracy.