Modeling of Lithology in South Pars Gas Field Using Artificial Neural Network

Coring from several wells, regardless of the oil field acreage, is an inevitable and necessary task in order to obtain general information of the region. Yet, coring in wells of huge fields is excessively costly. Therefore, finding a solution to avoid this excessive expense seems to be crucial. This work presents a type of artificial neural network modeling in order to use well bore logs in lithology prediction in one of the South Pars gas field reservoirs. Here, a network with three-layer back propagation (BP) method and Levenberg-Marquwardt algorithm has been used for lithology estimation. The network utilized density, neutron, gamma-ray and photoelectric effect (PEF) logs as inputs. Data from four wells in South Pars field has been used. Data from two wells (wells SPF1 and SPF2) having core analysis were used as network training, validation and testing. The network was then utilized to estimate the lithology in the two other wells (wells SPF3 and SPF4) and the results were compared with the core data (real lithology). The interval under investigation consists of Dolostone, Limestone, Dolomitic Limestone, Limy Dolostone, Anhydrite, Shale, Shaly Limestone and Shaly Dolostone. The mean square error (MSE) of rock types were 0.081 and 0.094 for SPF3 and SPF4 wells, respectively.