Research On Elastic Full-waveform Inversion Based On Step-by-step And Multi-scale Strategy

Full-waveform inversion makes the most of kinematic and dynamic characteristics of seismic data and can provide high-resolution estimation of subsurface elastic parameters.At present,full-waveform inversion has been extensively studied,but in the deficiency of low-frequency data,the inversion results are highly dependent on the initial model,and in the face of elastic media,the cross-talk between P-wave velocity and S-wave velocity leads to a strong multi-solution problem.Therefore,this dissertation carries out research based on the combination of"multi-scale"and"step-by-step"inversion strategies to reduce the dependence of full-waveform inversion method on the initial model and the multi-solution of the multi-parameter inversion results.This work can provide reliable elastic parameters for later oil and gas interpretation and is very important for seismic imaging,interpretation,and oil and gas exploration and development.The multi-scale inversion strategy of seismic data helps to reduce the nonlinearity of full-waveform inversion and the dependence on the initial model.However,in general,low-frequency information is not available,which greatly weakens the applicability of the multi-scale strategy.Therefore,it is necessary to make up for the lack of low-wavenumber components through other ways.Envelope inversion can extract ultra-low frequency information from seismic records,but this part of information is mainly concentrated in direct waves and refracted waves passing through the shallow and middle layer of the model.Therefore,the combination of envelope inversion and multi-scale strategy can reduce the dependence of inversion results on the shallow and middle regions of the initial model,but the dependence on the deep regions is still strong.In order to alleviate the dependence of the inversion results on the initial model,reflection full-waveform inversion is proposed to provide a better initial model for full-waveform inversion using the travel time information carried in the reflected wave,and then combined with the multi-scale inversion strategy to provide the estimation of subsurface velocity parameters.In reflection full-waveform inversion,the parameters characterizing the underground medium are changed from velocity to background velocity and reflectivity,which aggravates the problem of multiple solutions.In order to reduce the multiplicity of solutions,on the basis of applying the smoothing constraint along the orientation information of interface to the background velocity,the position information of the interface is obtained from the reflectivity model,and the background velocity is divided into interface and stratum interior,so the adaptive smoothing constraint perpendicular to the direction of the interface is applied to realize reflection full-waveform based on reflectivity-velocity consistency constraint.Reflection full-waveform inversion extracts the low-wavenumber update information of the model from the reflected wave,but discards the contribution of refracted wave.In order to make full use of the reflection and refraction information,Poynting-vector method is used to calculate the imaging angle in the imaging domain with less computational cost.The difference of the imaging angle between the low-wavenumber component and the high-wavenumber component is used to separate them.The low-wavenumber model update components from the reflected wave and the refraction wave can be obtained,and the joint waveform inversion based on angle domain filtering can be realized.In order to reduce the crosstalk between P-wave velocity and S-wave velocity,elastic full-waveform inversion based on multi-scale step-by-step strategy is studied.Specifically,the decoupled elastic wave equation is used to realize the wave field separation of P-wave and S-wave,and the P-wave velocity and S-wave velocity are inverted step by step according to the sensitivity difference of seismic data in the radiation patterns.In order to reduce the dependence of elastic full-waveform inversion on the initial model,the reflection full-waveform inversion of"pseudo acoustic data"using the hydrophone seismic data is used to improve the initial P-wave velocity model,and then the initial S-wave velocity model is obtained from the empirical formula.The numerical results show that the multi-scale step-by-step decoupling strategy can effectively reduce the dependence of elastic full-waveform inversion on the initial model,reduce the interference of P-wave velocity on S-wave velocity,and improve the inversion results of S-wave velocity.After obtaining an accurate velocity model,high-precision migration imaging can be carried out on this basis.In order to improve the resolution of imaging results,the least-square reverse time migration with sparse norm constraint is studied.Specifically,this dissertation compares the sparsity characteristics of different norms,uses the weighted L_1-2norm to represent the sparsity of imaging results,and the migration imaging results have higher resolution.