Abstract
Reservoir engineers tend to define a simulation model as "complex" when they have several of the following characteristics:
Complex fluids schemes loaded with multi PVT's or EOS's data sets associated with lateral and vertical compositional changes that might cause massive fluids mobility and blends, which resulted in unpredictable production trends.
Presence of tilted fluid contacts, buffer/tarmat zones, or other feature which increase the number of uncertain parameters and features. Each of which needs to be investigated individually to identify their impacts on field/wells levels history match profiles.
Limited sources of data available with high levels of geological heterogeneity with multi-layers and patchy zones, which result in creating many thin layers that would tend to impact dynamic model performance.
This paper is going to discuss the implementation steps and procedures utilized to history matching one of Abu Dhabi's complex onshore reservoirs model. This reservoir has high levels of unknown parameters with uncertainty. Our key objective is to achieve acceptable history match profiles which would be used for field development and optimization works. The second objective is to do this "quickly" in a more reasonable timeframe than is usually achieved.
In this complex model case study, it would be attractive to use Assisted History Match technologies to create a smart loop which could assess and modify multi-parameters with a linkage between static and dynamic models. This might require several applications to deal with both static and dynamic models in parallel. Using smart workflows all the possible geological parameters subject for uncertainty analysis have been identified and scripted directly in the dynamic model data deck which accelerates parameter adjustment; this allows us to carry out hundreds of runs in a more time efficient and practical ways using one tool. Although this study started with thousands of uncertain parameters, we reduced the unknown parameters into two main groups (A) Global parameters that would influences overall model results, such as shape of Kr's curves per rock types, and (B) Region parameters that would influences only the selected unit and/or region, for example faults transmissibility. After completing the 1st pass runs to optimize global parameters, those parameters were frozen, and the 2nd pass runs were started for the region parameters. This loop is continued for several iterations until most of parameters stabilized. In order to get maximum benefits from the existing CPU's, it was decided to coarsen the dynamic model while adjusting grid cells properties without mislaying geological features to accelerate runs at the beginning of this project.
By using smart workflows and a powerful AHM tool, it was easy to rank the converted models and to pick up the fewest representative cases that showed the lowest Objective Function errors. Several development scenarios were created, and these were used to assess multiple options, including: well objectives, configurations, and target locations with a single model. We then replicated that study on the rest of the converted models in order to check overall development risks envelopes (to assess production sustainability, pressure preserving, field max/min ranges of WCT and GOR, etc.….).
This paper presents our successful workflows, which have been utilized for the first time to compile static and dynamic models variables under one platform i.e. (introduce both static and dynamic models variables through dynamic data deck file, including to modify oil saturation option around history match problematic wells by redefining their rock types per region(s) while preserving on match quality with Sw_log profiles), as results we have managed to generate several cases that show good outputs while reflecting wide uncertainty ranges.
