Pre-stack Seismic Porosity Prediction Based On Bidirectional GRU And Attention Mechanism

Accurate prediction of reservoir parameters is a prerequisite for accurately describing reservoir characteristics and conducting reservoir evaluation,and plays a fundamental role in oil and gas field exploration and development.Rock porosity is one of the critical parameters to characterize the reservoir.High-precision prediction of porosity is conducive to more detailed descriptions of the location of highly porous and permeable reservoirs.However,the relationship between seismic elastic parameters and reservoir parameters is relatively complex,which brings some difficulties for accurate prediction of reservoir parameters.Traditional reservoir prediction methods have the problems of low efficiency and low accuracy.Deep learning technology has got broad application with its strong nonlinear extraction ability and high-speed parallel computing ability,providing a new idea for high-precision prediction of seismic reservoir parameters.In view of the above situation,this thesis applies deep learning technology to reservoir prediction and proposes a pre-stack seismic porosity prediction method based on bi-directional gated cyclic unit neural network and attention mechanism(Bi GRUAttention).The bi-directional GRU is used to realize the bidirectional propagation of information and the Attention mechanism is added to amplify the key information.The Bi GRU-Attention network is trained and tested using the P-wave velocity and density information obtained from pre stack simultaneous inversion of the borehole side seismic trace as input,and the real log porosity value as a label.The complex mapping relationship between seismic elastic parameters and porosity is established,and the network parameters are continuously optimized to obtain the best model,so as to achieve high-precision porosity prediction.The actual data test results indicate that compared with the conventional multiple linear regression(MLR),dense neural network(DNN)and gated cyclic unit neural network(GRU),the proposed Bi GRU-Attention network has higher prediction accuracy in blind well testing.The root-mean-square error(RMSE)of prediction results and logging data is less than 0.0022,and the mean absolute error(MAE)is less than 0.0014.Applying this method to the seismic data of a real 3D work area,the predicted porosity value matches well with the logging porosity value,which shows that this method has good practical value.