Multi-source Elastic Full Waveform Inversion Based On P-wave And S-wave Separation

Velocity is the core parameter in seismic data processing process.High-precision velocity modeling method plays a decisive role in migration,lithology,lithofacies interpretation and reservoir prediction.Elastic full waveform inversion(EFWI)has the potential to obtain accurate P-wave and S-wave velocity and density.Therefore,exploration geophysics is becoming more and more popular.This paper introduces the basic EFWI complete implementation process and formula.By comparing and analyzing the model test,the inversion accuracy of P-S wave velocity can be seriously disturbed by multi-parameter coupling.In the case of low precision of traditional EFWI,the method of full-wave inversion(SEFWI)based on vector P-wave and S-wave separation is proposed to alleviate the crosstalk effect.By analyzing the cause of crosstalk noise and the processing method of reference elastic wave reverse time migration(RTM),the P-wave velocity gradient and S-wave velocity gradient are calculated by PP wave field and PS wave field respectively.The realization process of SEFWI and the calculation method of its adjoint wave field are given.In order to get the step size accurately and avoid the interference of Pwave and S-wave,pure P-wave and S-wave seismic data are extracted directly in frequencywavenumber domain.Then,using pure P-wave target function and pure S-wave target function respectively,the iteration step of P-wave and S-wave velocity is calculated by three-point parabolic interpolation method.The high cost of computation and storage of EFWI is the main reason that restricts its large-scale application in industry.Although the local optimization method has greatly reduced the computational storage cost,the computational complexity is still huge,and the computational complexity of SEFWI is about twice that of EFWI.Therefore,the source coding technology is used to effectively improve the computational efficiency,and the four-variable dynamic source random coding technology is used to suppress the crosstalk noise generated by multiple sources.Model tests show that the full waveform inversion of elastic waves from multiple sources(ES?SEFWI)based on P-S wave separation can effectively improve the computational efficiency and inversion accuracy.Although the combination of four-variable dynamic source random coding technology and P-S wave separation technology can effectively suppress crosstalk noise,it still produces distinct discrete crosstalk noise for models with large interlayer range.And we know that underground velocity usually contains piecewise constant structures with sharp interfaces,while EFWI usually produces smooth inversion results.Therefore,the total variation regularization constraint(TV)is used to further suppress the discrete crosstalk noise and sharpen the boundary.In conclusion,we propose a strategy of full-wave inversion(TV?ES?SEFWI)for multisource elastic waves with P-S wave separation under the constraints of TV regularization.The model test and analysis of the ground observation mode and VSP observation mode prove that the proposed method and strategy are practical and effective.