Multi attribute transform and neural network in porosity estimation of an offshore oil field — A case study

This study focuses on one of the marine oil fields in the Southwest Iran. The main reservoir unit in the interested area is composed of an alternation of thin dolomite and anhydrite layers. In the available 3D seismic data, these layers cannot be resolved. Because of a composite detectable seismic response, it is difficult to make reservoir characterization by conventional seismic attribute analysis. ition to resolving the major reservoir rock units, the main aim of this study is to predict accurate porosity and make 3D porosity cube. For this purpose, at first, the 3D seismic volume was inverted to obtain acoustic impedance cube. In the next step, the acoustic impedance attribute besides other attributes extracted from seismic volume was analyzed by multiple attribute regression and neural networks to predict porosity. These linear or non-linear combinations of attributes for porosity prediction result in improved match between the actual porosity and predicted (in comparison with using only single attribute to predict porosity). An error of 1.1% in porosity prediction causes a change of 1500000 STBO in oil reserves in the oil reservoir. This shows the importance of using a better prediction method. imate the reliability of the derived multi attribute transforms, cross validation is used; according to the results it is found that the cross correlation between actual porosity and predicted porosity increased from 80% in the case of using a single attribute to 88% in the other case using multiple regression transform. Also, neural networks provide higher cross correlation values than both previous cases. Finally, according to the cross validation results, multiple regression transform is used for porosity prediction. implemented estimation technique, porosity slices prepared from the producing rock unit (A3 layer) provided a reliable result from lateral and vertical heterogeneities inside the A3 layer.