Reservoir Rock Type Identification Using Gustafson Kessel Algorithm and LOLIMOT Estimator

Identifying reservoir rock types or a petrophysical rock type is separating a reservoir to homogeneous part based on petrophysical characterization like porosity and permeability. There are different ways to study the reservoir or petrophysical rock types like; Winland’s equation, Pittman’s equation and rock quality index. For detecting and separating reservoir zones based on these properties, we need core data but in most wells even if in all parts of one well, there is no way to coring. Nevertheless, logging will do in all wells. Because of that, finding some ways to estimate porosity and permeability using other available properties is very important. In this assay, for studying the efficiency of the local linear model tree (LOLIMOT) neuro-fuzzy system in detecting petrophysical rock types, we compare the achieved result with the core analysis and MLP neural network algorithm. Based on the results, the local linear model tree (LOLIMOT) neuro-fuzzy system has great sufficiency in reservoir zoning because this method uses a divided and conquer strategy. Besides, this for decreasing the efficiency of this method and studying the effect of homogenization of data on the result, the data were clustered with Gustafson Kessel clustering method before using the neuro-fuzzy system. According to the results, zoning petroleum reserviors will be done based on petrophsical rock type with higher accuracy. Finally, in this study, based on Winland’s equation and using well logs and core analyses, 2 wells in one of the reservoirs in south part of Iran were clustered, and thereby zoning and petrophysical rock types were well detected and separated from each other.