Characterization and prediction of channelized systems is significant for the exploration of hydrocarbon and reliable estimation of reservoir production. However, the precise prediction of the spatial distribution of its heterogeneity, is a challenging task in the inter-well region because of irregular channel patterns. With the accessibility of 3D or 4D seismic data, there has been great interest in using machine learning models for seismic reservoir characterization such as facies classification and determining the nonlinear relationship between reservoir properties and seismic attributes. This paper will mainly focus on the application of neural networks algorithm to predict the presence of different seismic facies in heterogeneous channelized reservoir.
Supervised artificial neural natural (ANN) algorithms is a robust classification tool that integrates multiple seismic attributes into a number of classes to recognize seismic facies. In the first scenario, based on the available data, core and logging data analysis is conducted on main carbonate reservoir zone to recognize the development phases of the channel. In the second scenario, for a target carbonate tidal channel facies, we considered an approaches that used to classify and predict seismic lithology. We build and train Probabilistic Neural Network (PNN) to predict quantitatively the facies of carbonate channelized zone and the passageways laterally. We further quantified the distribution of tidal channel characterization through an acoustic impedance volume and seismically derived porosity volume. The representativeness of the channelized facies was analysed through blind tests using well logs as well as core data.
The most significant findings of our work that the 3-D seismic facies classification maps generated from PNN using appropriate seismic attributes provides two different facies which identified in the channelized systems. These facies are channels and restricted lagoonal related to depositional phases and flooding surfaces intervals. Our work clearly demonstrates that a multi-step workflow is important for classification and characterization of heterogeneous channelized systems facies. The value of this study is also to provide an interpretation of complex carbonate tidal channels, with quantification of the impact on dynamic behavior in the main oil-bearing reservoir of the case study.