Fareed Abdulla, speaking at the launch meeting of the Society of Petroleum Engineers new Data driven analytics (PD2A) technical section last month described ADCO’s use of ‘top-down’ modeling for ‘making difficult reservoir management decisions.’ ADCO has been investigating D2A since 2001 and began at-scale use in 2005 with a well production optimization on the Asab field. Here, water injection to the lower reservoir tended to break-through to an upper zone and kill the well. A study on a surrogate reservoir model used pattern recognition to train an intelligent system to identify candidate wells. Neural nets and genetic optimization of production rates was very successful.
The technique was then applied to Bu-Hasa, a large, geologically complex field where similar techniques were used to build a comprehensive reservoir model that has provided history matches of static pressure, time-lapse saturation and production rates for all the wells in the asset. The model’s validity was tested by sub-setting the data and performing ‘blind’ history matches—omitting several years of production data and getting the surrogate model to ‘predict’ the missing data.
The technique of top-down modeling requires no knowledge of the underlying physical processes involved. The idea being that such information is in general either unavailable or incomplete. It is better to build a surrogate model that is derived empirically from the ensemble of available measured data. More on TDM from PD2A luminary and professor of Petroleum Engineering at W. Virginia University Shahab Mohaghegh and from the SPE PD2A.
This article originally appeared in Oil IT Journal 2012 Issue # 11.
For more information or to comment on this topic email here.