Research On High-resolution Time-frequency Approaches And Applications In Reservoir Prediction

Seismic data belongs to non-stationary signal,the time-frequency approaches are required to depict the relationship between the signal with time and frequency.As a tool for seismic data analysis,the time-frequency analysis is able to provide valuable information for reservoir prediction.In this paper,the high resolution time-frequency approaches and their applications are discussed based on the traditional time-frequency methods.When seismic waves propagate through the gas charged reservoir,the high frequency components are dramatically attenuated.The ‘low frequency shadow' will show up near gas reservoir in the seismic data.In the paper,the synchrosqueezing transform(SST)and empirical wavelet transform(EWT)are analyzed and the ‘low frequency shadow' detection approaches based on SST and EWT are presented.Compared with the continuous wavelet transform,the two methods,SST and EWT,show high time-frequency resolution and can effectively describe variable frequency contents in the non-stationary signal,which is helpful to detect these anomalies associated with gas saturation.The VMD(Variational Mode Decomposition)is a novel adaptive signal decomposition method,which can non-recursively decompose the signal in time domain into an ensemble of modes with different bandwidth and the modes collectively reproduce the input signal.The Teager-Kaiser operator(TKO)is a local non-linear energy calculation approach.The combination of VMD and TKO is beneficial to identify the strong amplitude anomalies caused by gas reservoir.In addition,the dispersion is associated with attenuation resulted from the hydrocarbon,therefore,the dispersion-dependent attribute can be used as a tool for hydrocarbon detection.In the paper,the combination of VMD and Smith and Gidlow expression related with AVO equations is employed to perform the frequency-dependent AVO inversion.The high time-frequency resolution of VMD contributes to characterize and describe these anomalies in details so as to improve the ability to detect hydrocarbon.Both of the ‘low frequency shadow' and frequency-dependent AVO inversion demand the high quality seismic data.In the paper,an effective noise attenuation method based the compressive sensing and curvelet transform is proposed and compared with f-x deconvolution.On the other hand,the reconstruction of seismic data with missing traces based on local random sampling and curvelet transform is performed in order to enhance the recovery precision and is applied to the seismic data with missing traces reconstruction,which includes shot gather and 2D & 3D sections.Besides,the seislet transform has been used to interpolate regularly missing traces if an accurate local slope map can be obtained.The dealiasing capability of such method highly depends on the accuracy of estimated local slope.To ensure good slope estimation,low-frequency components of the aliased seismic data are employed to estimate the local slope in an iterative manner.However,a large number of iterations are required and the slope estimation is not stable via iterations.A new way to obtain the slope estimation is presented,first the NMO velocity is estimated and then a velocity-slope transformation is utilized to get the optimal local slope.The presented approach effectively decreases the iterations and recovers regular missing seismic traces.These improvements mentioned above are helpful to the subsequent seismic data processing and interpretation.