Shear wave velocity (VS) is a vital prerequisite for rock geophysics. However, due to historical, cost, and technical reasons, the shear wave velocity of some wells is missing. To reduce the deviation of the description of underground oil and gas distribution, it is urgent to develop a high-precision neural network prediction method. In this paper, an attention module is designed to automatically calculate the weight of each part of the input value. Then, the weighted data are fed into the long short-term memory network to predict shear wave velocities. Numerical simulations demonstrate the efficacy of the proposed method, which achieves a significantly lower MAE of 38.89 compared to the LSTM network’s 45.35 in Well B. In addition, the relationship between network input length and prediction accuracy is further analyzed.

Shear wave velocity is of great significance for accurate seismic data description and fluid tracking. However, the measurement of shear wave velocity is difficult, requiring high-precision equipment and professional operators. Conventional logging velocity often does not fully include the effective value of shear wave velocity, which brings serious challenges to the accurate exploration of oil and gas in deep strata. To meet demands of the industry, it is necessary to synthesize accurate shear wave logging from relevant conventional logging. Since the LSTM network is unable to compute sample points in parallel and has insufficient nonlinearity caused by sequential calculation, an auto-weighted sequence module is designed in this paper to extract context and better predict shear wave velocity. It takes each logging curve with a fixed step length to use a trainable matrix for auto-weighted rearrangement, and then uses the ELU function to activate. After the calculation of N such modules, the shear wave velocity is predicted by the fully connected layer. Comparing the prediction results of different methods, the auto-weighted neural network can indeed significantly improve the prediction accuracy, obtaining the highest accuracy with 39.21 of MAE and 0.9860 of R². In addition, the proposal way is better than the LSTM network at different input lengths, with a significant improvement of about 5%. In conclusion, the proposed way can effectively extract the sequence information of the input data, thus it is an effective sequence modeling tool, which has good performance in the prediction of the shear wave.