The Study On Full Waveform Inversion Of Earthquakes Based On Energy Cross-correlation Theory

In view of the demand for domestic oil and gas resources and the existing reservoir areas,the focus of resource exploration has shifted to deep subsurface and complex tectonic areas,which has resulted in higher requirements for the accuracy and resolution of subsurface structures and formations.Seismic exploration has become one of the most important resource exploration methods due to its large detection depth and high resolution.The velocity information of the subsurface medium is a very important parameter in the seismic exploration process and high accuracy velocity modelling can significantly improve the quality of the final seismic imaging.Among the existing velocity modelling methods in seismic exploration,the full waveform inversion method has the highest accuracy and resolution.Full waveform inversion can make full use of the kinematic and kinetic information in pre-stack seismic data to reveal tectonic details and lithological parameters in a complex geological context.As it embodies a mapping from data to model space,it makes the convergence of the objective function a strongly non-linear problem.The local optimisation algorithm also makes the full waveform inversion strongly dependent on the initial model.When the initial model is poor,the full waveform inversion will experience a period jump.In addition,the absence of low frequency components in seismic data,amplitude differences between data and inaccurate estimation of source wavelets also increase the likelihood of period jumps in full-waveform inversion,which in turn affects the progress of full-waveform inversion in practical applications.Therefore,a key issue is how to use full-waveform inversion for high-precision velocity modelling in a stable manner without being able to evaluate the observed data accurately.In this paper,two methods are proposed to overcome the problems that full waveform inversion is prone to period jumps and difficult initial model construction under complex conditions.Firstly,a locally correlated full-waveform inversion method based on an energy objective function is proposed to address the problem of insufficient low-frequency data in seismic data and the difficulty of matching amplitude information between data due to the non-linear decreasing amplitude of seismic waves with time.Secondly,a source-independent energy cross-correlation full waveform inversion method is proposed to address the problems of missing low frequencies,inaccurate amplitude simulation and difficult estimation of source wavelets that often accompany seismic data.Both methods can achieve stable full-waveform inversion under uncertainty of complex seismic data information.The methods proposed and the results obtained in this paper are summarised as follows:(1)The amplitude information of seismic data is susceptible to the influence of pre-processing and multiple parameters in the medium,and errors in amplitude can interfere with the correct convergence of the full waveform inversion.In this paper,we start from statistical local amplitude information and use the normalized cross-correlation objective function to weaken the effect of amplitude differences between observed and simulated data.The local normalized cross-correlation operation reduces the sensitivity of the objective function to differences in amplitude information by counting and normalizing the amplitude information of a single channel of seismic data within a certain length of time window.In addition,low frequency information is often missing from the actual seismic data.Combining the energy objective function for low frequency reconstruction alleviates the occurrence of cycle skipping problem to a certain extent and reduces the initial model dependency.The normalized local energy cross-correlation full waveform inversion method proposed in this paper can invert the macroscopic large-scale structure in the subsurface velocity structure more accurately in the absence of low-frequency components in the seismic data and with large local amplitude errors,and provide a good initial model for subsequent inversions.(2)Inaccurate source wavelets can have a significant impact on the final results of the full waveform inversion,while insufficient low-frequency information and large amplitude differences can also increase the risk of cycle skipping in the full waveform inversion.In this paper,we propose a source-independent energy cross-correlation method for full-waveform inversion.Firstly,the folded wavefield of the simulated and observed data is used to assimilate and cancel the imprint of wavelets from different sources.Secondly,an energy objective function is introduced to reconstruct the low-frequency components of the seismic data for inversion of the subsurface long-wavelength velocity structure.Finally,the normalized cross-correlation wavefield between simulated and observed data is used to reduce the focus of the objective function on residual amplitudes,highlight the matching of phase information,and further increase the inversion’s non-dependence and stability on amplitudes.At the same time,a corresponding multi-scale inversion strategy is proposed.The source-independent energy cross-correlation full-waveform inversion method proposed in this paper can obtain more accurate background velocity information and gradually supplement the small-scale perturbation information of the velocity model under the complex conditions of missing low-frequency components of seismic data,inaccurate estimation of source wavelets and errors in amplitude information,and has certain noise immunity.