Seismic Data Pre-stack Denoising Method And Application

In seismic exploration, high fidelity, high resolution and high signal-to-noise ratio of seismic data processing is the important basis and guarantee of seismic data interpretation, oil and gas reservoir prediction. Among them, the denoising of seismic data processing is the priority of seismic data processing. In terms of seismic data denoising, predecessors have done a lot of work, especially many mature methods about after stack data denoising. But as the terrain of seismic exploration region is getting more and more complex, especially in complex mountainous seismic data, the source of noise is getting very complex. Noise has strong interference on the effective wave, which leads low signal-to-noise ratio of seismic data, and has serious impact on the appearance of seismic data. The focus of the seismic denoising processing is turning to prestack. on the basis of the thorough analysis of the source and characteristics of prestack seismic data noise, corresponding seismic data denoising method is studied, separation of noise and effective signal is made, in order to achieve the goal of seismic data prestack denoising.In this paper, the causes of noise, development degree, type and characteristics of the seismic prestack record is analyzed, and the commonly used seismic data denoising method is introduced. the noise processed by denoising method is mainly surface waves, linear interference and random noise in the paper,.In this paper, on the basis of the study of conventional seismic data denoising, in terms of the situation of the low S/N ratio of mountain information, the low-frequency surface wave field and higher frequency effective wave field is separated with wavelet transform, and the coexisting wave field of the low-frequency wave and surface wave field is processed through F-K filtering, then low-frequency effective wave is picked up from it together with other higher frequency effective wave signals, finally the seismic wave field without surface wave is gotten, which achieve the separation of the surface wave. Compared with using f-k filtering denoising separately, the effect of using this method to remove surface wave is better.By studying the characteristics of random noise, linear moveout correction for bending events in seismic data is made.By using KL transformation, many components are acquired, each two of them are orthogonal and uncorrelated, the energy of coherent signal concentrates on the first few components, while the energy of noise distributes on all main components.Then reconstruct the main components which contains the effective signal, finally make inverse linear moveout correction to achieve random noise denoising.The noise is not continuous, unpredictable, which brought great difficulties to the denoising, through the analysis we can get the reflect wave event has spatial predictability in frequency-space domain, random noise is unpredictable, design to calculate prediction filtering operator on each frequency, then take the prediction filtering operator and the corresponding spatial direction data series make certain calculations, the signal can be predicted, for the bending events, the quasi-linear transformation can be used firstly, then do the processing, this achieve linear noise denoising, and it also have certain effect on the attenuation of random noise.The noise and the effective reflection information are often mixed together, none of the denoising method can achieve the fidelity of the data after denoising, so it is not be allowed that in order to complete inpursuit of high S/N radio, denoising too much and lose the effective signal.