The use of computational intelligence techniques has become increasingly popular in the field of geosciences. Artificial intelligence models are extremely useful for reservoir characterization, which requires high accuracy prediction for a good exploitation of the oil and gas resources. Porosity is one of the most important rock properties for fluid volume prediction and modeling of a petroleum reservoir. This study presents an improved approach utilizing the integration of intelligent systems to predict porosity. Firstly, the image analysis techniques were used for extracting petrographic parameters from thin section images. Consequently, the extracted parameters were input to two intelligent models, including a neural network (NN) and a fuzzy logic (FL) model. To improve the accuracy of the predicted porosity from NN and FL a genetic algorithm (GA) is employed, which combine the outputs of individual models. This methodology is known as committee machine with intelligent models (CMIS). The proposed methodology was applied to South Pars gas field, which is the largest non-associated gas accumulation in the world. The MSEs of the NN and FL methods for prediction of porosity in the test data are 1.49×10-2 and 2.21×10-2, which correspond to the R2 values of 81.6 and 84.6, respectively. The MSE of the CMIS for the test data is 1.09×10-2, which corresponds to the R2 value of 87.6. It is clear that, the CMIS performed better than NN and FL models acting alone.
Developing a Committee Machine Model for Predicting Reservoir Porosity From Image Analysis of Thin Sections
Rostami, Amirshahriar , Hatampour, Amir , Amiri, Morteza , Ghiasi-Freez, Javad , and Mehdi Heidari. "Developing a Committee Machine Model for Predicting Reservoir Porosity From Image Analysis of Thin Sections." Paper presented at the SPWLA 20th Formation Evaluation Symposium of Japan, Chiba, Japan, October 2014.
Download citation file: