Study Of Seismic Noise Suppression Methods Based On EMD And SVD

With the continuous deepening of oil and gas exploration,the focus of exploration has gradually shifted from conventional and structural oil and gas reservoirs to unconventional and lithological oil and gas reservoirs.The complex exploration environment makes the collected seismic data contain various types of noise.In order to have a good understanding of subsurface geological information,it is necessary to improve the signal-to-noise ratio of seismic data.This paper researches a series of denoising methods based on Empirical Mode Decomposition(EMD)for suppressing random noise in seismic data.First of all,aiming at the problem of endpoint effect in EMD,three endpoint processing methods including mirror symmetric extension,boundary local feature scale extension and polynomial fitting are proposed.Through comparison,it was found that mirror symmetric extension and boundary local feature scale extension have better effect.Secondly,a series of improved EMD methods were proposed to address the modal aliasing problem of the EMD methods,including Ensemble Empirical Mode Decomposition(EEMD),Complete Ensemble Empirical Mode Decomposition(CEEMD),and Complete Ensemble Empirical Mode Decomposition Adaptive Noise(CEEMDAN).These methods were used to suppress noise in the constructed seismic signals,and it was found that the CEEMDAN method has the best denoising performance,but there is still a problem of incomplete noise suppression.Therefore,this paper proposes a denoising method(CEEMDANSVD)that combines EMD and singular value decomposition(SVD).This method filters the intrinsic mode function(IMF)and remaining terms obtained from CEEMDAN through correlation coefficients,and uses SVD to suppress noise for IMF with high correlation coefficients,then completing signal reconstruction.Using this method to denoise the constructed seismic signal,it was found that this method suppresses noise more thoroughly than the CEEMDAN method.Finally,the above method was applied to denoise synthetic seismic records and actual seismic data,and it was found that CEEMDAN-SVD has the best denoising performance and the highest signal-to-noise ratio of the data.