Fig 1.
General diagram of shear wave predicting.
Table 1.
the main structural differences among different algorithms.
Fig 2.
ANN structure diagram.
Fig 3.
LSTM network cell internal structure.
Fig 4.
LSTM network overall structure.
Fig 5.
diagram of the GAN-LSTM network.
Note: In the Fig 4, ,
,
are the set of network parameters of feature extractor, the generator and the discriminator respectively;
and
are loss functions for the generator and the discriminator.
Fig 6.
structure of the GAN-LSTM network.
Table 2.
Statistical characteristics of variables of two wells.
Fig 7.
Scatter diagram of logging curves in Well A.
Fig 8.
Scatter diagram of logging curves in Well B.
Fig 9.
Correlation of logging curves with shear wave velocity.
Fig 10.
Logging sequence conversion.
Table 3.
The GAN-LSTM network parameters.
Fig 11.
Flowchart of the algorithm.
Table 4.
Feature vectors generated by LSTM and GAN-LSTM.
Fig 12.
Predictive results of three neural networks in well B.
Table 5.
Network deployment environment and training time.
Fig 13.
Local amplification comparison in Well B (the upper part marked in
Fig 12).
Fig 14.
Local amplification comparison in Well B (the lower part marked in
Fig 12).
Fig 15.
Comparison of the probability distribution of the three prediction methods.
Fig 16.
The relative error of the three forecasting methods.
Table 6.
Comparison of accuracy under different methods in two wells.
Fig 17.
MAE different methods in well B.
Fig 18.
Cross plots of different methods.
Fig 19.
Fig 20.
Prediction of LSTM and proposed GAN-LSTM network under different input in Well B.
(Input length from left to right is set as 5, 7, 9, 11). (Mark the upper and lower parts of the curve in red for comparison).
Fig 21.
Local magnification of LSTM and GANLSTM’s prediction under different input length.
(upper part in Fig 20). (The input length of (a), (b), (c), (d) is set as 5,7,9,11, respectively).
Fig 22.
Local magnification of LSTM and GANLSTM’s prediction under different input length.
(lower part in Fig 20). (The input length of (a), (b), (c), (d) is set as 5,7,9,11, respectively).
Table 7.
Evaluation of prediction in Well B under different input window length.
Fig 23.
Comparison of R2 of results between LSTM and GANLSTM.
Fig 24.
Comparison of MAE of results between LSTM and GANLSTM.