Prediction of residual water saturation using genetically focused neural nets

The "genetic petrophysics" approach for predicting petrophysical parameters using genetically focused neural nets (GFNN) has only recently been introduced. The approach only requires a minimum number of core plugs, along with associated wireline log data, from a chosen representative genetic unit (RGU). This case study has successfully developed and tested this new methodology to predict residual water saturation. Combinations of wireline logs and core data from a short 7m RGU interval in a North sea well were used to train the GFNN predictors. These were then applied to predict the residual Sw throughout the whole logged section in the training well and adjacent wells in the same field. Traditional hydraulic unit analysis provided the basis for selecting the minimal training plugs. Only 4 core plugs were finally required to represent the hydraulic units in the RGU and provide good results. This approach is very cost effective in terms of core material and computing time. Presently we have only tested this approach in oil bearing shoreface reservoirs. Thus, it is recommended that this approach be tested in other environments. Copyright 2004, Society of Petroleum Engineers Inc.