Ghobad Ravanan
Jamshid Moghadasi
Mohammad Ali Mohammadi - Corresponding Author’s Address: Department of petroleum engineering ,Omidiyeh Branch ,Islamic Azad University , Omidiyeh ,Iran

Ensemble Combination Artificial Neural Networks (ANN), a type of Committee Machine is in this study to estimate the effective porosity of reservoir rock was used. Petrophysical log data wells 1, 3, 6, 9 and 13 super giant South Pars field in the gas member k-1 and k-2 of the Kangan Formation was selected for the case study. Wells 1, 3 and 13 for training and wells No. 6 and 9 can be generalized to evaluate the networks go to work. Sonic, density, gamma and neutron logs, as input and effective porosity, as the output networks were considered. Performing a long trial and error stage, five three-layer networks with error back propagation training algorithm, which had the best generalization ability, combine to make the ensemble chosen. This collection of the best network, network structures 1 - 4 - 4 May, the stage is extended. The correlation coefficient of 98.38 percent and square root of the mean square error was 1.2930. Using a simple averaging methods and algorithms MSE-OLC, 26 may be combined ensemble collection network 5, were constructed and their results were compared with results of the best single network. Ensemble combining the best combination of network No. 1, 2, 4 and 5 the MSE-OLC method is extended in stages.The correlation coefficient of 98.55 percent and square root of the mean square error was 0.1 in 2343. Thus, the combined ensemble, RMS estimates for data validation, decreased 4.54 percent.

Ensemble Combination, Artificial Neural Networks (ANN),Kangan Formation, Back propagation(BP), mean square error (RMS)