Improved Fractured Reservoir Characterization Using Neural Networks, Geomechanics and 3-D Seismic

Abstract

Abstract In this paper, a new approach which uses seismic attributes in a quantitative manner to enhance the characterization of fractured reservoirs is presented. The new approach uses the seismic travel time to identify the reservoir structure and the thickness of the fractured producing formation. Using these data, a quantitative geomechanical model is constructed. When comparing the geomechanical models derived from seismic data and mapping methods, it becomes apparent that many structure details may be misrepresented and/or missed when interpolation methods are used for defining reservoir structure. Using a Neural Network and the available well data, the geomechanical model is correlated with the oil production. This model is compared to the seismic amplitude which appears to provide the best indication of fracture intensity in the case of the field studied. Given the seismic data, which are available over the entire reservoir, and the fracture model found by the neural network, the overall reservoir fracture network is predicted.