Reservoir Characterization: A Machine Learning Approach

‘Reservoir Characterization (RC)’ can be defined as the act of building a reservoir model that incorporates all the characteristics of the reservoir that are pertinent to its ability to store hydrocarbons and also to produce them. It is a difficult problem due to non-linear and heterogeneous subsurface properties and associated with a number of complex tasks such as data fusion, data mining, formulation of the knowledge base, and handling of the uncertainty. This present work describes the development of algorithms to obtain the functional relationships between predictor seismic attributes and target lithological properties. Seismic attributes are available over a study area with lower vertical resolution. Conversely, well logs and lithological properties are available only at specific well locations in a study area with high vertical resolution. If a functional relationship can be calibrated between seismic signals and lithological properties at available well locations, then distribution of these properties across the study area can be predicted from available seismic information. Depending on the distribution of the lithological properties, a dataset can be classified into two categories – balanced and imbalanced. Sand fraction, which represents per unit sand volume within the rock, has a balanced distribution between zero to unity. On the other hand, water saturation, oil saturation etc. has an imbalanced distribution skewed at one and zero respectively. The investigation about the sand fraction (balanced distribution) variation over the study area has been attempted as a prediction problem; whereas, the distribution of water saturation (balanced distribution) has been approached as a classification (Class low/ Class high) problem in this work. The thesis addresses the issues of handling the information content mismatch between predictor and target variables and proposes regularization of target property prior to building a prediction model. In this thesis, two Artificial Neural Network (ANN) based frameworks are proposed to model sand fraction from multiple seismic attributes without and with well tops information respectively. The performances of the frameworks are quantified in terms of Correlation Coefficient (CC), Root Mean Square Error (RMSE), Absolute Error Mean (AEM), etc. After successful completion of sand fraction prediction, a one-class classification framework based on Support Vector Data Description (SVDD) is proposed to classify water saturation from well logs. The designed framework is modified to include seismic variables as predictor attributes to obtain the variation of water saturation over the study area. In other words, the class labels (Class low/Class high) of water saturation belonging to a well location can be predicted from seismic attributes by the modified classification based framework. The proposed frameworks have outperformed other supervised classification algorithms in terms of g-metric means and program execution time (in seconds).