Using information entropy theory and bayesian decision method to identify appropriate parameters for evaluating and discriminating oil facies (mansuri oil field, south of Iran)

Due to subsurface heterogeneity and existing vagueness in geophysical interpretation, identifying and interpretation of facies in wellbores is always prone to uncertainty and risk. Nowadays several methods have developed for quantitative facies interpretation. These methods are generally divided into deterministic and stochastic categories. Deterministic methods, in spite of their simple modeling procedure, cannot expose the amount of error or accuracy of the model. On the other hand, stochastic methods, in addition to quantifying the error of the model, can provide the probability of the model’s accuracy in each point of the reservoir. The Bayesian approach is one of the stochastic methods that use conditional probabilities for modeling. This approach, as well as probabilistic modeling of hydrocarbon facies, quantitatively computes the effect of additional data in decreasing the error of the classification. Information entropy theory, by quantifying the intrinsic uncertainty in each model input parameter, can easily provide the selection of valuable parameters. The present study was carried out on one of the wells of Mansuri oil field, south of Iran. After generation of training data by using rock physics techniques and Gassmann’s relation, the value of each input parameter was identified by entropy analysis. Then, by use of Bayesian analysis and valuable parameters, oil facies classification and discrimination was implemented. The five optimum parameters were elastic impedance, compressional wave velocity, shear wave velocity, density and porosity. The amount of error in this method is approximated to be 11 percent. This investigation also showed that gamma ray parameter does not have a drastic positive effect on identification and discrimination procedure of oil facies, which has a good agreement with the results of entropy analysis.