Determination of water saturation by angular competitive neural network

Abstract The accuracy on determining water saturation is decisive to perform a realistic evaluation of hydrocarbon reserves and critical to reduce the economic risks in the oil industry investments. Water saturation is the solution of the Archie equation, but in many situations, a quick and convenient solution of this equation may be a hard problem, mainly when there is no confidence on porosity and when the conventional core analysis does not supply the cementation exponent and the water resistivity. In these cases, an association of the Hingle plot with the Pickett plot may be used to solve the Archie equation, but as any graphic these methods are frequently subject to visual misinterpretation. In this work, we assume the resistivity–porosity dependence in the Archie equation as an angular pattern and introduce an intelligent algorithm based on a new model of artificial neural network, named as angular competitive neural network, which was designed to discover angular patterns present in the data. This characteristic makes the angular competitive neural network be able to produce all parameters needed to solve the Archie equation, performing angular pattern recognition in the Hingle plot and in the Pickett plot. This method is presented with synthetic data and evaluated using actual wireline logs and conventional core analysis.