Prediction of Water Saturation by FSVM using Well Logs in a Gas Field

Water saturation is one of the key petrophysical parameters that mainly affects the accuracy of initial oil estimation related to a hydrocarbon reservoir. Approximation of this parameter is inevitable since it has a high effect on economic development of hydrocarbon reservoirs. In this paper, we approximate a function, using two wells with two core data sets belonging to each well, to predict water saturation by means of Support Vector Machine (SVM) algorithm in one of the gas reservoirs in the Persian Gulf. Due to the inevitability of noise and outliers in the measured data, SVM is modified to Fuzzy SVM (FSVM). For this purpose, a membership function is applied on the points, so each data point receives a membership degree. In this case, each input point is able to contribute to the learning of decision function. In other words, FSVM is able to enhance SVM by devoting less value to noise and outliers, as a result, better models compared to SVM can be produced. In this study, application of SVM for regression purpose (Support Vector Regression) is carried out on eight logs of DT, GR, RHOB, NPHI, LLD, LLS, MSFL, PEF as input with relevant core data belonging to a gas zone. Then, we determine the coefficients based on the comparison between predicted water saturation (using both SVR and fuzzy SVR algorithm) and core data. Our results show that the predicted water saturation from fuzzy SVR and SVR are 95% and 71%, respectively (higher for fuzzy SVR than SVR).