Simulation of high resolution reservoir models is useful to gain insight into oil and gas reservoirs. Nowadays, massive comprehensive reservoir simulation models can be built with detailed geological and well log data. These models require a very large high performance computing (HPC) platform for conducting reservoir simulation. Saudi Aramco has developed a state-of-the-art simulator, GigaPOWERS™, which is capable of simulating multibillion cell reservoir models. This paper provides an overview of challenges related to constructing, HPCs and visualizing the simulation output of giant reservoir models, and how the computational platform at Saudi Aramco is designed to overcome these challenges. A large HPC platform can be designed for reservoir simulation by connecting multiple Linux clusters in a simulation grid. Such an environment can provide the necessary capacity and computational power to solve multi-billion cell reservoir models. Such a simulation grid for reservoir simulation has been designed in the Saudi Aramco EXPEC Computer Center. In this study, we provide the benchmark results of multiple giant fields to evaluate the performance of the Saudi Aramco simulation grid for reservoir simulation. Communication and I/O routines in the simulator may add a considerable overhead in computation on such a computing platform. Connectivity between clusters on our simulation grid is tuned to maintain a high level of scalability in simulation. Excellent scalability results have been obtained for computations of giant simulation models on the simulation grid. Simulation models of the order of one billion cells pose a challenge to pre- and post-processing applications to load and process data in a reasonable time. Remote visualization, Level of Detail and Load on Demand algorithms were implemented in these applications and data formats were revised to efficiently process and visualize massive simulation models.

Identifying potential productive reservoir units for infill drilling is a major challenge in developing giant fields in order to meet production targets and extend plateau. A common practice in identifying potential drilling locations is using oil saturation from production logs or near-by wells performance. Literature already recommended combining static and dynamic parameters from reservoir models (geological or numerical) to calculate cell performance indices such as Reservoir Opportunity Index (ROI) or Simulation Opportunity Index (SOI). It is difficult for those methodologies to provide volumetric representation of the hydrocarbons in potential areas since they do not specify the details of the clustering mechanism of similarly performing cells. The proposed algorithm allows the reservoir engineer to, progressively and recursively, define hydrocarbon sweet spots areas. In this work, progressive-recursive self-organizing maps (PR-SOM) is developed and tested on a carbonate reservoir model. PR-SOM is an unsupervised artificial intelligence neural network algorithm that classifies the simulation grid cells into potential drilling targets by using a progressive list of dynamic or static reservoir parameters to identify similarly "good" contiguous regions. This is achieved by applying SOM on geomodels based on relative permeability, fluid saturation, pressure, and displacing fluid influx in the first iteration. The second iteration, PR-SOM explores the already selected regions and applies SOM on mobile and immobile hydrocarbons. The last iteration recursively applies PR-SOM to identify areas away from existing wells on the already defined regions from last iteration. The algorithm allows further definition of sweet spots based on more parameters and will further increase the potential value of the classified regions. PR-SOM was applied to a carbonate reservoir with the objective of identifying un-swept areas as potential candidates for infill drilling. To compare the resulting potential target regions, an implementation of sweet spot identification algorithm and conventional approaches were applied on the same field. The results show that PR-SOM generated more accurate and conservative regions reducing the risks and increasing the confidence in the designated regions. In addition to obtaining more accurate clustering results, using PR-SOM allows extending the search to increase the value of required targets whereas previous work has to adhere to the original selected parameters (static or dynamic) for region identification and selection.

Modern reservoir simulation models include very detailed description of geology to enable accurate physics solutions. In order to increase the reliability of the simulation forecast, simulation models grew in size to a billion or more grid blocks, which poses a challenging complexity to build, validate, and history match the models. To help assess and handle new complexity and inherent uncertainties, an artificial intelligence algorithm based on Self-Organizing Maps (SOM) has been developed to explore and identify geologic model components based on similarities and dissimilarities in the model. Sector models can be generated based on these region definitions for focused region assessment, history matching, production/injection optimization, or optimal well configuration. In this work, an unsupervised artificial intelligence neural network algorithm is devised that can combine both static properties distribution (permeability, Reservoir Quality Index (RQI), or porosity) and dynamic properties (pressures or saturations) to construct a similarities matrix. Standard deviation-based chaos minimization is then applied to merge similar regions (contiguous), identify similar but distant regions, and reduce the number of identified heterogeneous regions to the desired controlled number. The algorithm was applied on a synthetic reservoir model using the GigaPOWERS® simulator, The developed SOM-based algorithm conditions geology to reservoir dynamics, and reduce model computational requirements of assisted history matching (AHM) with minimal engineering effort.