In Alberta there are many carbonate pools with large vertical relief havingsignificant gas caps or solvent zones above a diminishing oil leg. Productivityforecasting of the oil wells as gas or solvent coning becomes dominant, is ofprime concern. This paper describes a technique to develop a two parameterconing correlation from available production data, via a plotting procedure.The nature of the correlation lends itself to the development of an independentcrossplot which checks the consistency of the correlation with observed wellperformance. In addition, the difficulties often associated with GORcorrelation development, i.e. resolving rapid GAR change with minor changes inoil rate, can be overcome with this new approach.
Once developed, the correlation can in turn be used with either conventionalmaterial balance models or coarsely gridded reservoir simulation models topredict future performance. Examples from the Westerose D-3 Pool will be usedto illustrate important aspects of the correlation development and productionforecasting procedure.
There are many high relief carbonate reef pools throughout Alberta which arereaching the final stages of oil production. Historically, the pools have beenundergoing gravity stable miscible or immiscible displacement throughout theirproductive lives. For many of these pools the remaining oil sandwich hasreached the state, where coning of gas, solvent and/or water is unavoidable.There is still a need to optimize oil production from the remaining sandwichprior to these pools being placed on blowdown to recover the remaining solventor gas.
The situation where significant coning takes place has always lead to problemswhen field production forecasts are necessary. This is as a result of thereservoir simulation tools available to engineers. Typically, one of twoapproaches has been undertaken. The first involves performing single wellconing studies for each well or group of wells and somehow extrapolatingperformance to the field. The weakness of this approach is the the lack ofinteraction of the single well model with the rest of the reservoir.
A second approach is to model the system with a conventional large rectilineargrid system. Since these models cannot properly account for the localized nearwellbore drawdowns, separate single well radial models are used to generatevertical pseudo-relative permeability curves. These curves are then used asinput to the larger model. This is a time consuming process which involves aconsiderable number of runs with both sets of models with no guarantees ofsuccess. In practice this second approach has not been used due to thedifficulties of separating coning behavior from overall reservoir performance.An added complication is the need to continually re-run the single well radialflow model to update the vertical pseudo-relative permeability curvesthroughout the history matching process.
Recently, advances in state of the art for reservoir simulation have been madewhich allow for local grid refinement around wells2,3,4. In this waythe entire study can be run from one model without recourse to artificialdevices. However, no practical guidelines for the degree of local gridrefinement necessary to adequately model coning are available in theliterature.