COMPUTATIONAL METHODS FOR RESERVOIR CHARACTERIZATION IN STUDYING THE EFFICIENCY OF THE TERTIARY MIGRATION IN ABANDONED OIL RESERVOIRS

The tertiary migration of hydrocarbons in abandoned oil reservoirs, after a primary production stage, is a spontaneous process that, in some reservoirs, leads to a new reformation of the saturation state inside the collector rock which becomes economically attractive and, thus, making the production resumption potential very high at, practically, low costs. To study the efficiency of such a process, data regarding the geological model, physical model and production model are needed. Due to the fact that the well's dimensions, compared to the reservoir dimension, is negligible, the values of some properties (collector rock properties, from the physical model) can be considered isolated. Flow simulations are based on large grids with a big number of cells that need to be populated with values for the necessary parameters, values that are needed for the area between the wells, where these parameters can only be estimated. This paper discusses two computational methods, neuro-fuzzy networks and polynomial regression, that are proposed for estimating two of the collector rock parameters (porosity and initial water saturation), for the zones between the wells. Also, these methods are applied to a hypothetical reservoir in which pronounced heterogeneity of these parameters is present and results are presented that show which is the best way to evaluate the needed parameters between the wells when heterogeneities are present in collector rocks.