Effective Information Recovery Of Seismic Date Based On Adaptive Filter Technology

Seismic exploration has always been an important means of exploring fossil energy like oil and natural gas,in order to learn more about the structure and distribution and study the underground structure in detail,improving the accuracy and resolution become emphasis and trend in the development of seismic exploration.With the gradual development of exploration degree and the application of 3D wide azimuth seismic exploration method,it is more difficult to acquire the complete seismic data volume.Because of the restriction of actual environment,economic factors and some other conditions,it is very difficult for observation system record complicated information of seismic wave field.Seismic data always show the spatial discontinuity in the process of acquisition,on the one hand,because of the complexity of the actual exploration environment,the presence of obstacles such as village,river causes the change of the observation system;on the other hand,because of the restriction of economic factors,detectors and shot points cannot be arranged continuously,these factors cause the lack of seismic data.But complex seismic data are the precondition of many important processing methods,such as surface related multiple wave elimination,wave equation migration,time lapse seismic and 4D seismic exploration and so on.Predictive filter can effectively express the energy spectrum of seismic data,which provides an effective approximate estimation method for solving the problem of geophysical inversion and also provides a tool for recovering the missing seismic information.Prediction error filter is essentially the autoregressive of signal,which is applicative in the space-time domain and frequency-space domain.Seismic data essentially is nonstationary,the conventional seismic data processing methods tend to have higher requirements for seismic data,like the hypothesis that seismic event is stable plane wave.In recent years,many geophysicists discuss the problem of multiple waves from different perspectives.On the one hand,multiple waves is regarded as noise and removed from the seismic data,on the other hand,in seismic data processing,it is a forefront direction that how to better use the underground reflection information which is contained in multiple waves.In this paper,we use this theory as the core and apply the predictable character of kinetics of seismic data,and then through the adaptive filter technology,we start from the path relationship between primary wave and multiple waves to solve the recovery question of seismic effective information with the interference of internal multiple and in the case of missing data.Firstly,this paper reviews the classical theory of prediction filter,in order to characterize the non-stationary characteristics of seismic data better,and then raises the theories of the adaptive prediction error filter and streaming prediction filter.In the process of reconstructing the missing seismic data,the lack of seismic traces leads to inaccuracy of filter estimation.Therefore this paper proposes to use the cross-correlation of multiple wave and primary wave to build virtual primary wave,and does interpolation reconstruction for missing near offset data according to the wave field information of data themselves.Because there are amplitude and phase differences between the primary wave which is built through multiple wave and real primary wave,we propose to represent virtual primary wave energy spectrum which is based on adaptive prediction error filter of non-stationary autoregressive process,and use the least-square inversion method to reconstruct missing data of near offset,adaptive prediction error filter realizes local adaptive characteristics by solving mathematical underdetermined problems under the restriction of regularization.This method possesses high interpolation accuracy,but the speed of calculation is slower.This method is suitable for the processing of complex seismic fields.We propose a technical plan which is based on the combination of streaming prediction filter virtual primary wave to reconstruct missing seismic data.In the process of applying streaming prediction filter to estimate the filter coefficients,we use the proportional similarity of adjacent filter coefficients as constraint condition,accomplish filter estimation quickly and efficiently without multiple iterations,not only improve the computational efficiency,but also reduce the computational memory,predict missing seismic data through the kinetic information which is contained in multiple wave of data themselves and provide basis of data analysis for more reasonable filter estimation in the process of data interpolation.This plan fulfills the fast construction of large-scale data in the case of losing definite interpolation accuracy.This plan is suitable for the seismic field which has relatively complex,low accuracy,but a huge amount of data.The problem of internal multiples has been the difficulty in the field of seismic exploration.With the deep development of seismic exploration,lithologic exploration and structural reservoir are gradually transitioning to complex reservoir,such as unconventional,convert and lithological and so on.In addition,the requirement for the precision of seismic imaging is continually improved.In reflection seismic data,the multiple wave is often regarded as unwanted noise,because it often interfere with the effective signal imaging of primary reflection,so it is suppressed in seismic data processing.Internal multiples have been ignored by people because they have weaker energy and are not easy to identify,but with the development of deep oil and gas exploration,the problem of internal multiples are gradually broke through.Due to the limitation of the traditional denoising method,it is difficult to effectively suppress the internal multiples.In this paper,we first separate the raw data and then use the cross-correlation technology to construct virtual event.The virtual event exists as a formation of non-causal wave and provides data basis for predicting internal multiples.The method that suppresses internal multiples by using virtual event belongs to the predicting subtraction technology of wave equation.The predictive results of the internal multiples often has deviation with the original data on the amplitude and time of arrival,so matching subtraction technology plays an important role on final results.Nonstationary adaptive prediction filter gets rid of the hypothesis of overlapping window and better characterize the non-stationary of earthquake.In this paper,we propose a non-stationary adaptive prediction filter method which is based on the plastic regularization.We use this method to conduct adaptive matching for internal multiples,which make the matching multiple wave energy keep pace with the multiple wave energy of original data and achieve the purpose of suppression,and fulfill the recovery of effective information of non-stationary seismic data with the interference of internal multiples.The processing results of actual data shows that the adaptive filter technology can characterize the non-stationary of seismic field effectively.Through the kinetics relationship of seismic field,the technology perfects the transformational relationship between primary wave and multiple waves and offers effective basic instrument for solving the recovery(including the interpolation of seismic data and the suppression of internal multiples)problems of effective information of seismic data.Finally,we form a set of complex technological tactics.