عنوان مقاله [English]
Estimating of various petrophysical parameters and determining of subsurface geology is very important in petroleum reservoir evaluation. Exploration drilling and various well logs normally provide this sort of information. Among the various well logs, the photoelectric (PEF) log is very important as it able to determine the lithology of the reservoir precisely. Therefore for those wells for which this log is not available it is necessary to predict them somehow. In this study it is aimed to use artificial neural network (ANN) ability to tackle this problem. To achieve the goals, a back-propagation ANN (BP-ANN) is planned to model the interrelationships between seven different well logs, and PEF logs. Data from three wells in the Ahvaz oil field (Asmari reservoir) are organized into training, testing and validation data sets for BP-ANN modeling. Data of the fourth and the fifth wells in the same field are retained as independent data sets for evaluating the ability of the network PEF prediction. Once the designed network has been trained properly, its performance has also been tested. When it has been found that the performance is satisfactory the data set of the fourth and the fifth wells are applied to the trained network. The results of the ANN modeling show that the designed network with three layers and architecture of 7-10-1 can produce the precise PEF log that compares well with the measured PEF logs. This means that the designed network is capable enough to predict the PEF logs for the required wells in the same area.