Implementing the teaching-learning based optimization algorithm to predict shear wave velocity from well logs in sandstone and carbonate case studies

Seismic wave velocity along with petrophysical data provide valuable information at the exploration and development phases of oil and gas fields. The compressional-wave velocity (Vp) is acquired using conventional acoustic logging tools in many drilled well. But the shear-wave velocity (Vs) is recorded using advanced logging tools only in limited number of wells, mainly because of the high operational costs. So, alternative methods are often used to estimate Vs. Heretofore, several empirical correlations which predict Vs by using well logging measurements and petrophysical data are proposed. But these empirical relations can only be used in limited cases. The use of intelligent systems is an efficient approach for predicting Vs. In this study, in addition to the modified Greenberg-Castagna method, we used the teaching-learning based optimization (TLBO) algorithm to make linear and nonlinear models for predicting Vs. This algorithm is used to make prediction in a sandstone formation from an offshore oil field located at Western Australia and a carbonate formation from an onshore oil filed located at south west of Iran. We compared the estimated Vs values using TLBO algorithm with observed Vs and also with those predicted by modified Greenberg-Castagna relation. The results are showing the algorithm efficiency. The results of linear and nonlinear models are also very close together, but the difference is that the linear model is faster than the nonlinear model and is preferred for predicting Vs. Using the linear model shows that for the sandstone formation the error percent is 2.3 and the regression coefficient is 0.82 and these values are 3.3 percent and 0.95 for the carbonate formation respectively. These values show that the linear model of TLBO algorithm can be used as an efficient way to predict Vs.