Application of Artificial Neural Network for Assisting Seismic-Based Reservoir Characterization

Abstract

 Reservoir rock physical properties, such as porosity and water saturation, always play prominent roles in the development of oil and gas fields. Accurate information regarding their distribution is always desired. For this new approach that uses a purpose, a combination of intelligent computing (artificial neural network or ANN) and rock physics, with a full utilization of core data, well logs and seismic-derived attributes, is proposed. The method is basically an effort to link the required rock physical properties to seismic- derived attributes through the use of rock physics theories. The ANN itself is used to fill the gaps of data array required by the proposed method through its capacity for pattern recognition. The proposed method is applied to a limestone reservoir in East Java. Validation is carried out by comparing the results to the observed data at well locations as well as by geological justification. The application has shown a new potential for supporting reservoir modeling and field development.