Using Data Driven Analytical Models with Compositional Simulation in Doubling the LPG Production from Bahrain Field

摘要

Crestal gas injection started in the Bahrain Field in 1938 and since then, 1,900 Bscf has been injected in Mauddud, the main oil producing reservoir in the Bahrain Field, creating a secondary gas cap. Furthermore, since 1965, an estimated 100 MMstb of Liquefied Petroleum Gas (LPG), or 14% of the bypassed oil, has been recovered from the secondary gas cap through stripping with the remaining oil volume in the secondary gas cap being approximately 700 MMstb. Today, the remaining oil saturation in the Mauddud gas cap is estimated to be approximately 40%. This paper presents the results of an extensive study that was made to forecast the Bahrain Field associated gas compositions and potential gas-liquids production recovery. Several forecast methodologies were used including data-driven analytical models, a compositional cross-section model, and a full-field compositional history matched model. The results of these forecasts and the conclusions are presented and compared. In this study, two scenarios of different gas compositions of injected gas and their impact on gas plant liquids recovery are explored. In addition, this paper addresses the challenges and uncertainties associated in forecasting the gas compositions and ways to overcome them. The data-driven models and compositional cross-section model were initially used, however, due to their inherent uncertainties, a full field compositional simulation model was necessary. This compositional model was history matched with a seven (7) component Equation of State (EOS) to capture the lighter hydrocarbon components. Moreover, this model was used in predicting the yield and composition of the existing gas recovery plant. The results from all methods recommend doubling the capacity of the existing plant, which was commissioned in late 2018. A comparative analysis found that data-driven models can be used for gas cycling when using the same gas injection compositions. However, data-driven models over-estimate the Liquefied Petroleum Gas (LPG) yield if leaner gas is used for gas injection, which is the case for the proposed gas plant expansion.