ASSISTED HISTORY MATCHING OF EARLY FIELD LIFE AND PROBABILISTIC FORECASTING USING AN INTEGRATED SUBSURFACE - SURFACE NETWORK NUMERICAL MODEL: BANGKA DEEPWATER DEVELOPMENT

摘要

Indonesia Deepwater Development (IDD) is the first Chevron ultra-deepwater development project in Indonesia. The first field development, Bangka Field, was put on production in August 2016. Bangka is developed by two subsea wells, each producing from individual stacked gas condensate reservoirs. Production is processed at the West Seno FPU via a single flow line. Bangka Field consists of two deepwater upper slope channel reservoirs. The structural model was developed using a Pre-Stack Depth Migration volume integrated with petrophysical data from exploratory and production wells. A reservoir simulator is utilized to model compositional fluid flow in the reservoir and wells up to the sea floor. The network of subsea flow lines, manifold and riser are modeled with a pipe flow simulator. Coupling of both model types improves forecast reliability. Probabilistic history matching commenced within a month of first gas to decrease uncertainty in the production forecast. A Design of Experiments-based assisted history matching workflow was developed to rapidly screen out multiple Earth models. The short history duration resulted in short model run times and allowed a large number of geologic models to be explored. As the historical production record was extended with each month of production, fewer models passed screening criteria. Additionally, an iterative workflow was employed where simulation-based geologic learnings were passed back to the Earth modeler for revision of the subsurface characterization before re-development of the simulation model. Following each round of history matching, all models were applied in a base production forecast to simulate metrics key to differentiating low- from high-side production outcomes. For business purposes, the asset team applied plateau time and EUR, at both the reservoir and field level, to differentiate models. Discrete model selection was then performed to quantitatively select P10, P50, etc., forecast models. This work demonstrated that utilizing early-production data can be employed to rapidly reduce subsurface parameter uncertainty.