ABSTRACT: Shear wave velocity (Vs) and compressional wave velocity (Vp) play an extremely important role in identifying reservoir lithology, physical properties and fluids in geophysical exploration. Due to the lack of logging equipment or limited funds, data related to Vs isn’t available for all wells, especially old wells. Therefore, lots of methods have been developed to predict Vs using other available well logs. In this study, a series of logging data consisting of shear wave velocity (Vs), compressional wave velocity (Vp), neutron porosity (NPHI), resistivity (RDEP), caliper (CALI), spontaneous potential (SP), gamma ray (GR) and density (RHOB) were collected and then models for Vs prediction were developed using multi-variable linear regression (MLR), multi-variable polynomial regression (MPR), empirical models, deep neural network (DNN) and Random Forest respectively. Results show that the accuracy of the models firstly increases to the peak value and then decreases with the increase of input logging parameters. The prediction result of Random Forest is more accurate and stable compared to other models. However, the performance of DNN model was disappointing, even worse than multiple regression in terms of stability. This shows that the prediction of Vs may not achieve higher accuracy and stability with complex machine learning model such as DNN, some simple machine learning models may perform better because of the strong correlation between well logs.

This content is only available via PDF.
You can access this article if you purchase or spend a download.