Distributed Unstructured Grid Infrastructure for Complex Reservoir Simulation

Abstract

It is now standard practice to use reservoir simulation as an everyday tool for better engineering and management of oil and gas fields. Modern reservoir simulation must be comprehensive in the treatment of complex reservoir processes, complex geology, intelligent maximum reservoir contact (MRC) wells, and production systems. For the more mature giant oil fields, significant volumes of detailed data are available. It is critical to maximize the use of these data for field planning in order to optimize recovery. Nowadays fast inexpensive parallel computers are widely available, modern reservoir simulators can exploit parallelism effectively such that modeling can be done at the desired scale and speed for better and timelier reservoir management. Simulators must also have the flexibility to use the general structured or unstructured grid systems for any combination of complex geology and recovery processes. To handle all these complexities in a massively parallel environment, we have developed in Saudi Aramco's in-house reservoir simulator POWERS a distributed unstructured grid infrastructure (DUGI). DUGI is the parallel data structure and algorithms which constitutes the simulator's nerve center. Typical complex models for giant fields can have tens to hundreds of millions of grid cells, handling hundreds to thousands of wells, including some complex MRC wells. It can contain geologic features such as faults, pinch-outs, and fracture and multimodal porosities, described as the multiple porosity multiple permeability (MPMP) system. Both structured grids, with irregular multilevel local grid refinements (LGR) and unstructured grids can be used to resolve local reservoir or well geometry, and to model dynamic near-well coning behaviors. DUGI unifies the management of all these complexities into a consistent, scalable, and extendable system. It is designed to be modular such that data complexity and communication management are hidden from the rest of the simulator. This enhances simulator extensibility for ongoing feature development. Elements of DUGI will be discussed and examples of simulation models and results provided.