RESERVOIR CHARACTERIZATION USING ARTIFICIAL NEURAL NETWORK IN X FIELD, NIGER DELTA, NIGERIA

Abstract

An integrated approach to reservoir characterization involving seismic attributes extraction and Articial Neural Network (ANN) analysis of the reservoirs of X eld, onshore, Niger Delta was carried out to assess the effectiveness of ANN as a tool for hydrocarbon reservoir study. ANN is a relatively new technique and imitation of the human brain in its basic form. In this study, it was used in the prediction and classication of reservoir properties and facies from well logs and seismic. Facies classication on logs was executed using an empirical relationship between selected logs such as gamma ray (GR), density (DEN) and resistivity (RES) logs which were cross-plotted against one another to determine data suitability. Facies classication on seismic was employed to predict facies distribution without well control. Two attributes, Root Mean Square (RMS) and Relative Acoustic Impedance (RAI) were selected based on their capability to discriminate lithologies. Facies classication on logs showed correlation between GR and DEN, GR and RES, DEN and RES logs to be 69%, 35% and 36% respectively. These values fell within the acceptable range. Facies classication on seismic revealed 44% correlation between RMS and RAI. Hence, ANN analysis effectively distinguished reservoir sands from nonreservoir sands and accurately identied lithologies penetrated by the wells of the Field. The unsupervised neural network was able to distinguish water and hydrocarbon-bearing sands. This technique had proven to be an effective tool for facies distribution studies and could be employed for generation of leads and prospects for hydrocarbon exploration.