Optimization Of Full-waveform Inversion Based On The First-order Velocity-stress Acoustic Wave Equation

Compared with conventional oil and gas,unconventional oil and gas have the disadvantages of low reserves,small scope,and difficulty in exploration.At the same time,conventional data processing methods have limited processing capabilities and poor processing results.We need to adopt new processing methods to improve the seismic data processing capability under complex structures.Compared with the reverse time migration,the full waveform inversion can directly invert the velocity parameters of the model medium,and the inversion results are more accurate,which can facilitate the subsequent interpretation and analysis.However,full waveform inversion has the problems of low calculation efficiency,large amount of calculation,and excessive memory usage,which seriously restricts the further promotion and application.This thesis takes the two-dimensional first-order velocity-stress equation as the starting point,first deduces the forward and reverse time migration algorithms,and then deduces the full waveform inversion algorithm.For the large amount of calculation and storage occupancy,useing pseudo-depth optimization method and the optimization algorithm was programmed using the C++AMP parallel computing architecture.In the forward and inversion,the accuracy of the numerical simulation is mainly affected by the number of model grids,but an excessively high number of grids will increase the amount of calculation.It is very important to reduce the number of grids while ensuring the complete acquisition of model information.The pseudo-depth method is to transform the acoustic wave equation and velocity field in the conventional Cartesian coordinate system into a curved coordinate system and carry out forward inversion operation.This method can reduce the computation and storage occupation by reducing the number of longitudinal grids without losing the model information.This method can match the wavelength with the velocity field so that the wavelength can be sampled uniformly in all velocity layers.Further,in order to improve computing efficiency,the algorithm is implemented by C++AMP parallel computing architecture,which is more common than at present As far as GPU parallel architecture is concerned,it has the advantages of simple programming language and environment,good hardware adaptability and so on.Through the trial calculation analysis of the international standard model and the actual block model,the applicability of the pseudo-depth domain full waveform inversion algorithm is verified.After numerical simulation experiments,found that:Compared with the CPU algorithm,the C++AMP parallel computing architecture used in this thesis can bring about an acceleration ratio of about 40-80,which greatly improves the efficiency of forward and inversion calculations,and the application of the pseudo-depth domain method can additionally.The application of the pseudo-depth domain method can bring about an additional 20%-30% reduction in storage occupancy and about 10%-20% time-consuming reduction.The large time-consuming and memory-saving reduction makes the full waveform inversion method more suitable for production.