| With the development of geophysical exploration of the oil and gas,the objective of the deep seismic exploration,especially in guidelines of development about hydrocarbon,will be one of the most important,as be used to accomplish the task of increasing O il and Gas reserves.Generally,the geological condition(middle or shallow depth)of the subsurface maybe more friendly than that in the deep layer,as the seismic data coming from deep layers have low signal-noise ratio,the internal multiple which has stronger energy covers the primary one and the assume of the common stacking hyperbolic equation can’t be applied in practice.Moreover,the energy of the seismic date from deep layers is too weak to be recognized and be easily eliminated as the random noise.In this paper,we proposed an application of spectral decomposition using regularized non-stationary autoregression(SDRNAR)to recognize the weak primary from deep layers.This method aims at decomposing the signal of deep seismic data into several spectral components,each of which has a smoothly variable frequency and smoothly variable amplitude.This method can suppress noise data w hen decomposing and reconstructing original signal,on the basis of which can broaden target spectrum of target spectrum using Gabor deconvolution method.Seen from the result of model data and field seismic data,the deep seismic resolution have been improved.The interformational multiples of deep seismic data is one of the most important factor in affecting imaging effectively.Conventional method severely rely on model data,and this will remain interformational multiples which will cause pseudomorph.I n the paper we will take the advantage of mathematical morphology to separate primary wave and multiples.Seen from the application of model data and field seismic data,we get a better result.As the geological conditions becomes more complex,we cannot get accurate estimation of deep layer velocity,so the event cannot match conventional stacking hyperbolic form,and this will cause weaken of far offset information in stacking.The CRS(the common reflection surface)method do not rely on velocity,so we can replace the stacking velocity with reflection bin location,inclination and curvature to realize CRB stacking.In the paper,we improve C RS.Firstly using similar weighting stacking along the offset on pre-stack gathers,and then using plane-wave flatting method along event to stack.Seen from the application of model data and deep field seismic data,the weak event energy in deep layer have been strengthened.In the process of propagation,the attenuation of frequency and amplitude will occur because of the geometric diffusion,completely inelastic case and etc,especially in deep seismic data.We use quality factor Q-value to express absorbing and attenuation function of formations.To realize the absorption by estimating the Q-value,we need variable assumptions.In this paper,we applied morphology multi-scale algorithm on deep seismic data to compensate the absorption and attenuation,and then compensate frequency by deconvolution.This method needs no assumptions and has a high adaptability in deep seismic compensation of absorption and attenuation. |
