Full text loading...
-
Soft-Computing Methodology for Prediction of Water Saturation in Fractured Carbonate Reservoirs
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 80th EAGE Conference and Exhibition 2018, Jun 2018, Volume 2018, p.1 - 5
Abstract
This paper describes a methodology integrating different soft computing techniques to predict petrophysical properties based on well log data. One and two hidden layers feed-forward neural network (FFNN), as well as support vector machine prediction models are compared. K-medoids clustering is further used to improve the prediction results. Well coordinates (only for the FFNN model), well depth and six wireline logs, commonly used in evaluating petrophysical properties, Gamma Ray (GR), Formation Density (RHOB), Neutron Porosity (NPHI), Sonic Travel Time (DT), Deep Laterolog (LLD), and Caliper (CALI) are used as input data. A total of 84.5 m of core with available data of water saturation completed with the saturation values obtained by the best petrophysical models are used for calibration. Four different scenarios are considered for model testing including that of a well unseen by the model. The proposed approach is evaluated using a real-world data set from the naturally fractured oilfield asset located in the coastal swamps of the Gulf of Mexico. The results show a strong correlation between the given data and those obtained from soft computing methods (R= 98% for training and R=90% for testing on a new well).