Shear Velocity Prediction in the Tight Oil Formation with Deep Learning

Shear velocity is an important parameter for reservoir characterization especially in the unconventional reservoirs such as tight oil and shale gas, but shear velocity log is not available in the field for its high cost, so it is crucial to predict shear wave velocity with conventional logging data. The existing methods of shear wave prediction mainly include empirical formula and rock physics modeling. The empirical formula method is inappropriate for predicting shear wave velocity in the occasion of complex lithology, and many parameters need to be optimized in the process of rock physics modeling, thus it was difficult to be applied widely. This paper focuses on the predicting shear wave velocity with deep feedforward neural network (DFNN) subsequent to the quality control and preprocessing of the input logging data, this method was proved to be much more effective in predicting the S-wave velocity through validation wells than empirical and rock physics modeling method.