Modelagem totalmente implícita de poços na simulação composicional de reservatórios de petróleo

Abstract

This study aims to implement and evaluate the effects of implicit well modeling in fully implicit compositional reservoir simulation, considering different grid types, including structured grids—Cartesian and corner-point—and unstructured grids, combined with the finite volume method. The fully implicit modeling of production wells was implemented and tested in the UTCOMPRS compositional simulator, developed at the University of Texas at Austin, using the Peng-Robinson equation of state to describe hydrocarbon phase behavior. The study accounted for the interaction between wells and the reservoir by integrating the well equations with the reservoir equations. This approach allowed wells to operate not only under a prescribed oil flow rate but also under a prescribed total liquid flow rate, while also enhancing the physical accuracy of production rate estimations and improving computational efficiency compared to the semi-implicit model (the original simulator model), particularly when refined grids were used. Unlike traditional methodologies in UTCOMPRS, which use pressure, phase saturation, mole numbers, and bottom-hole pressure from the previous iteration to calculate well production rates, the proposed approach evaluates the well simultaneously with the reservoir variables. This enables the consideration of the well- reservoir pressure gradient and the fluid column weight along the completions. The Newton-Raphson method was employed to solve the nonlinear system resulting from this fully implicit formulation. The results indicated a significant reduction in the number of iterations, particularly for an anisotropic reservoir using a corner-point grid with 14 production wells, which achieved a 64.72% reduction in iterations compared to the semi-implicit model, leading to a 10.5-hour decrease in simulation time. Additionally, a maximum absolute average relative deviation of 1.48% was observed in gas flow rate compared to the commercial simulator used. Finally, the results highlight the importance of the well-reservoir pressure gradient in flow rate estimation, especially in scenarios with multiple wells and heterogeneous permeability fields.