Reservoir zone prediction using logging data - multi well based on levenberg-marquardt method

Well logging is a well-known and effective method for oil and natural gas exploration in new fields in order to enhance oil and gas production. Well Logging is defined as an acquisition method to qualitatively and quantitatively evaluate the existence of hydrocarbon layer in the well. In this research, we studied the relations between well logging data and reservoir zone in Salawati basin, Irian Jaya area. Four well logs with four attributes such as Log Gamma Ray (GR), Log Resistivity (ILD), Log Density (RHOB), and Log Neutron (NPHI) were explored. The reservoir zone data has been previously determined by using log curve whether it is a reservoir zone or not. This data then is being used as a target for learning. Since the logging data is a complex and nonlinear, Levenberg-Marquardt (LM) was then implemented as an artificial intelligent algorithm in performing this study. The objective of this work is to build decision support system that will automatically find reservoir zone based on well logging data. The results of this work showed that Mean Absolute Percentage Error (MAPE) of training for reservoir zone prediction by exploiting Levenberg - Marquardt is 0.3803 % with 500 iteration. Validity test results based on ROC curve with cross validation folds 10 is 84.9984% and area of under ROC is 0.992. This result showed that this method has a high potential to be used in real exploration activities so that the predicting reservoir zone then can be done precisely.