Improvement in Seismic Reservoir Characterization by Artificial Intelligence in Heterogeneous Media

Abstract

For any reservoir engineering issue or manage production from the petroleum reservoir, it is required to have seismic characterizations in quantitative manner, rather than qualitative geological interpretations. Herewith, seismic inversion could assist reservoir engineer as the technique to transform seismic data to quantitative rock properties. General steps in interpretation of seismic data preparing for porosity estimations consist of seismic structural interpretation, inversion procedure and attributes analysis. Since there is no direct measurement for the lithological parameters, they are to be computed from other geophysical logs or seismic attributes. This process also requires repeated intervention of the experts for fine tuning the prediction results. Standard regression methods are not suitable for this problem due to the high degree of the unknown nonlinearity. The problem is further complicated because of uncertainties associated with lithological units. In this context, Artificial Neural Network is considered to be useful tools to establish a mapping between lithological and well log properties. In this study, a strategy is presented for defining 3D seismic reservoir porosity model based on advanced method of artificial intelligence (AI) concept. This strategy then would be applied on a complex and heterogeneous oil reservoir which is a relatively symmetrical anticline whose trend is N-S. Required input data was prepared by seismic attribute and the velocity was modeled by vertical seismic profiling data. The general characterization strategy followed by initial inversion model construction for acoustic impedance of total cube for the target formation. Consequently, initial inversion model for effective and total porosity of the target formation was obtained. Acoustic impedance logs were used for neural network training and the genetic algorithm were used for calculation. High correlation values around 86% in cross plots, confirm accuracy of the porosity estimation by the AI method. This model then was used to precise the geological and geometrical properties of the reservoir for well location proposal.