A mass-conservative INSIM-FT data-driven model
Please login to view abstract download link
Simulation of oil reservoirs is a crucial activity in the oil industry, influencing major decisions in field exploitation strategies that involve millions or even billions of dollars. Various methods for modeling waterflooding in oil reservoirs have been developed, with conventional models utilizing grid-based full-scale reservoir numerical simulators. However, these models are computationally expensive, with a single run taking several hours. Consequently, data-driven models have emerged as effective alternatives to address this limitation. The Interwell Numerical Simulation Model with Front-Tracking (INSIM-FT) \cite{guo2018physics,guo2018waterflooding} is a hybrid, physics-based data-driven model that eliminates the need for detailed reservoir geological characterizations. Instead, it relies on injection and production data and general geological information. The reservoir is approximated by a network of 1D interwell connective flow units, each characterized by transmissibility and pore volume. The solution follows the implicit pressure/explicit saturation (IMPES) approach, solving pressure equations on well nodes and using the front-tracking algorithm to analytically solve the Buckley–Leverett equation for calculating fraction flow rates of each unit. The model is coupled with the ES-MDA method \cite{EMERICK20133}, which is an iterative data assimilation method. ES-MDA works by creating an ensemble of possible input parameter sets, called realizations, and then using a smoothing algorithm to generate a more accurate estimate of the true state of the reservoir. This study specifically focuses on investigating water and oil mass conservation in INSIM-FT. The goal is to understand observed errors in the mass balance of the simulations and propose a method for their reduction. Results demonstrate that the proposed method effectively reduces mass balance errors, enhancing the model's representation of the physics of the problem.
