An Evaluation of a Method to Predict Unknown Water Levels in Reservoirs and Quantifying the Uncertainty

Abstract

Abstract A technique developed by Hawkins, et.al.(l) which integrates log and core data to predict water levels in hydrocarbon reservoirs has been evaluated using Alaska, Gulf of Mexico and North Sea reservoirs. The method is based on the construction of a synthetic capillary pressure curve from porosity, permeability, and water saturation at each available data point in the rock column. Regression analysis is used to estimate the Free Water Level (FWL) and an associated confidence interval. Conversely, if the FWL is known, the model can be used to estimate porosity, permeability, or water saturation, when two of the three parameters are known. Typically, water contacts are located by drilling down structure wells, by obtaining pressure information from RFT's or DST's, or by capillary pressure data. Unfortunately, estimating water contacts from RFT's, DST's and down structure wells requires the water contact be penetrated. Available capillary pressure measurements, although scarce, can be used to estimate the vertical distance to the FWL, defined as the point of zero capillary pressure for the hydrocarbon/water column. This is performed by superimposing the estimation of the water saturation, at that depth, on the curve. The distance to the FWL (or water contact) from that point can be identified by converting the capillary pressure curve to height above the FWL. Although penetration of the water contact is not required with this method, it does assume homogeneity of the rock column. However, the advantage of the capillary pressure model technique over the above technique is that it does not assume the rock interval is homogeneous but uses measured porosity, permeability, and water saturation data. Additional strengths of the capillary pressure model approach to determining water level are: reliable FWL estimation in a transition zone, prediction of rock properties and water saturations, and a consistency check of capillary pressure data. No analysis of data consistency was performed in this work. However, it is apparent that since the model is a functional form of capillary pressure behavior, estimates of displacement pressures can be checked along with the shape of the data in the high pressure region.