Artificial neural network (ANN) models are designed to emulate human information processing capabilities such as knowledge processing, speech, prediction and control. The ability of ANN systems to spontaneously learn from examples, reason over inexact and fuzzy data, and provide adequate responses to new information not previously seen, has generated increasing acceptance for this technology in the engineering field and resulted in numerous applications. A preliminary investigation into the use of this novel technology is presented towards predicting formation damage by quantifying wettability and two-phase relative permeability of oil reservoirs.

An artificial neural network model based on the Back Propagation technique is trained with a number of variables from experimentally established relative permeability (relperm) curves. The reservoir core input data covers an extensive range of porosities and permeabilities from different lithologies having diverse wettabilities. The trained model is then tested with only a couple of easily obtainable input variables such as the Swc and Sor and predictions are made on the wettability and relperm curves. A change or shift in the relperm curves is associated with changes in wettability, and perhaps to formation damage in the drilling process.

The wettabilities of the rock-fluid system are predicted to within 90 % of the experimentally determined values. The relperm curves, particularly the end-points are predicted to within 85 % of the measured results. The accuracy of the predictions are significantly enhanced with model training using more precise reservoir data and better defined formation lithologies. Neural networks have immense potential in predicting relperm curves and thereby assessing formation damage in reservoirs.

You can access this article if you purchase or spend a download.