Multicomponent seismic reservoir characterization of a steam-assisted gravity drainage (SAGD) heavy oil project, Athabasca oil sands, Alberta

Abstract

Steam-assisted gravity drainage (SAGD) is an in situ heavy oil recovery method involving the injection of steam in horizontal wells. Time-lapse seismic analysis over a SAGD project in the Athabasca oil sands deposit of Alberta reveals that the SAGD steam chamber has not developed uniformly. Core data confirm the presence of low permeability shale bodies within the reservoir. These shales can act as barriers and baffles to steam and limit production by prohibiting steam from accessing the full extent of the reservoir. Seismic data can be used to identify these shale breaks prior to siting new SAGD well pairs in order to optimize field development. To identify shale breaks in the study area, three types of seismic inversion and a probabilistic neural network prediction were performed. The predictive value of each result was evaluated by comparing the position of interpreted shales with the boundaries of the steam chamber determined through time-lapse analysis. The P-impedance result from post-stack inversion did not contain enough detail to be able to predict the vertical boundaries of the steam chamber but did show some predictive value in a spatial sense. P-impedance from pre-stack inversion exhibited some meaningful correlations with the steam chamber but was misleading in many crucial areas, particularly the lower reservoir. Density estimated through the application of a probabilistic neural network (PNN) trained using both PP and PS attributes identified shales most accurately. The interpreted shales from this result exhibit a strong relationship with the boundaries of the steam chamber, leading to the conclusion that the PNN method can be used to make predictions about steam chamber growth. In this study, reservoir characterization incorporating multicomponent seismic data demonstrated a high predictive value and could be useful in evaluating future well placement.