Simulation technology is constantly evolving to take advantage of the best available computational algorithms and computing hardware. A new technology is being jointly developed by an integrated energy company and a service company to provide a step change to reservoir simulator performance.

Multiscale methods have been rapidly developed over the last few years. Multiscale technology promises to improve simulation run time by an order of magnitude compared to current simulator performance in traditional reservoir engineering workflows. Following that trend, an integrated energy company and a service company have been working in collaboration on a multiscale algorithm that significantly increases performance of reservoir simulators.

The numerical accuracy of a reservoir simulation depends on the accuracy of each stage in the simulation algorithm. However, not all internal computations require the same amount of work to maintain stability and reach the necessary level of accuracy. In particular, pressure and transport calculations are governed by different physics and tend to have quite different numerical characteristics.

A multiscale black-oil reservoir simulation technology has been recently developed and implemented in a reservoir simulator used by the industry. It is based on a sequential fully implicit formulation. Compared to the regular fully implicit formulation, this formulation adds flexibility in the choice of solution strategy for reservoir pressure and transport. The reservoir pressure is computed by a multiscale algorithm that uses coarsening combined with reconstruction of fine-scale mass-conservative fluxes. The transport of fluid is computed using a Schwarz overlapping method based on the coarsening for the pressure solution. The multiscale method has proved to be accurate and reliable for large real-data models. The new solver is capable of solving very large models an order-of-magnitude faster than the current commercial version of the solver.

You can access this article if you purchase or spend a download.