Research On Random Noise Suppression Method Of Seismic Data

In recent years,global oil and gas demand has gradually increased.With breakthroughs in exploration technology,the exploitation of complex oil and gas reservoirs and unconventional oil and gas has become a new focus of oil and gas resource exploration.Seismic records produce a lot of random noise due to the complicated acquisition environment,which seriously affects the accurate interpretation and subsequent processing of seismic data.Therefore,need to effectively suppress the random noise in the seismic data and extract the weak effective signal from the low snr seismic data.The research object of this paper is the suppression of random noise in low snr land seismic exploration data and microseismic data.For the above two different seismic data types,time frequency peak filtering(TFPF)and Shearlet are developed respectively,In view of the deficiency of the two algorithms in random noise suppression,the corresponding improved schemes are put forward respectively,and through theoretical analysis and simulation test and actual data processing results,it is proved that the improved schemes have achieved good results in weak signal extraction and waveform protection.The TFPF algorithm is a relatively new time-frequency denoising method.The advantage of this algorithm is that it can recover non-stationary signals without any assumptions,and is widely used in one-dimensional signal processing.In recent years,TFPF algorithm is applied to the seismic signal processing field,used in the micro seismic data in random noise suppression,and achieved good results For one dimensional problem of random noise in seismic data,based on the analysis of traditional TFPF algorithm in seismic data of the insufficiency of the noise suppression,gives the corresponding improvement program,and the improved algorithm is adopted to improve the filtering.The influence of the filter window length on the TFPF algorithm is analyzed.The traditional TFPF filter uses a fixed window length,which cannot strike a balance between random noise suppression and waveform maintenance.According to the different distribution ranges of effective signal and random noise in the frequency domain,this paper uses complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)to decompose into intrinsic mode function(IMFs)in the frequency domain,and calculate the sample entropy(SE)of IMFs.The noisy signal is divided into a signal dominant part and a noise dominant part,and an adaptive window length adjustment scheme is constructed.The signal-dominant part adopts a shorter filter window length to retain the effective signal amplitude,and the noise-dominant part adopts a longer filter window length to better reduce noise,thereby improving the TFPF algorithm’s ability to respond to nonlinear one-dimensional microcomputers at low signal-to-noise ratios.The filtering accuracy of seismic signals.Due to the complex surface mining environment of complex oil and gas reservoirs,seismic data generally presents the characteristics of "weak signal,strong interference",which is difficult to process by conventional filtering methods,which limits the identification of effective signals.Therefore,this paper has carried out the research on the two-dimensional Shearlet transform random noise suppression technology,and constructed a restoration scheme suitable for effective signals in land seismic exploration data with low signal-to-noise ratio.Shearlet transform has strong sparse representation ability,localized features and directional sensitivity,can realize multi-scale and multi-directional decomposition of signals,and can reconstruct effective signals through sparse coefficients.In this paper,the classical Shearlet global threshold can not be adjusted according to the scale and direction of the problem,this paper carried out the Shearlet transform on the theoretical data with noise,according to the distribution of effective signals at different scales,and then calculated the Shearlet coefficient L2 norm in different directions at different scales,and rearranged the data,and analyzed the based on the distribution characteristics of the effective signal under the scale,an adaptive threshold denoising method which varies with the scale and direction is proposed to remove the random noise to the maximum extent and retain the effective signal.Two-dimensional Shearlet threshold denoising only processes single shot records without considering the correlation between common detection points.The difference is that the three-dimensional Shearlet transform transforms the multishot seismic records into the three-dimensional Shearlet domain,fully considering the correlation between the shots and representing the seismic records more sparsely,and then removing the random noise of multi-gun seismic data through thresholding.Taking into account the difference between the effective signal and the random noise distribution characteristics,based on the two-dimensional Shearlet transform denoising,combined with the adaptive threshold function to remove the seismic random noise,and then using the three-dimensional Shearlet inverse transform,the weak signal can be effectively restored and the snr can be improved.This paper studies the current shortcomings of two time-frequency denoising algorithms,TFPF and Shearlet transform,and constructs a corresponding optimized random noise suppression scheme,which overcomes the shortcomings of the two algorithms in seismic data processing and improves the denoising of traditional algorithms.performance.Realizes the reconstruction of weak signals under low signal-to-noise ratio,protects the edges and texture details of seismic data,and balances the maintenance of microseismic signal waveforms and noise suppression,providing reliability for improving the effect of oil and gas reservoir development and finely reconstructing the structure of oil and gas basins Technical guarantee.