Spatial modeling of oil exploration areas using multilayer perceptron neural network (MLP) in GIS

The exploration of hydrocarbon resources as a process is very complex and costly. In this process, multiple factors of geology, geochemistry and geophysics are prepared and combined together. Designing the best route to take seismic data and determine the best location for drilling exploration wells is extremely important because improper or careless determination of the location, time consuming and expensive during the operation. The aim of this study was to identify possible areas for oil and gas in the map of 1: 250,000 Ahvaz with 20 oil fields using multilayer perceptron neural network (MLP) and geographic information systems. For this purpose, 17 maps of factors including: the lowest and highest values (total organic carbon, potential for the production of hydrocarbons, peak Tmax, the production index, the oxygen index, the hydrogen index) and the proximity to areas of an high Bouguer gravity anomaly, an anticline axis and faults, map of the topography and the curvature of the yield curve Asmari subsurface were created by GIS functions. For the combined factor map, the multilayer perceptron neural network (MLP) that is data-driven methods was used. The validation results showed that the neural network 17×10×5 is better than the other models with a R = 0.8948 ,RMSE=0.0267 and the Kappa=9079. Besides, the neural network 17×10×5 is able to accurately predict the oil fields. On the other hands, Some fields could not be identified, and also, some areas were classified oil fields mistakenly.