Resolution-Oriented Weighted Stacking Algorithm Research

Horizontal stacking of seismic data is an important step in structural interpretation,lithology interpretation,and reservoir prediction,which is a bridge between the shotgather data and the underground structural data.The conventional horizontal stacking algorithm is based on the equal weight among traces in the pre-stack gather,that is,equal-weight stacking.Because of the low-pass filtering effect of stacking,the signal-to-noise ratio is improved while the seismic resolution is poor.Post-stack resolution-enhancement processing is usually required to improve the accuracy of interpretation and inversion.Absorption attenuation in pre-stack gathers related to seismic travel time and propagation path,or improper seismic processing methods(such as excessive ground-roll suppression and pre-stack deconvolution considering only near-offset data)will result in attenuation along the spatial direction of seismic gathers.With the increase of offset,the dominant frequency,and bandwidth decrease.Equal weight stacking will blur the frequency difference and waste the near-offset data with high resolution.Meanwhile,because of the low-pass filtering effect of stacking,the stacking resolution is usually low.Frequency distribution characteristics in pre-stack gathers are fully considered,and the resolution-oriented weighted stacking algorithm is adopted.Taking the stacked ’broad bandwidth’ and ’high dominant frequency’ as the processing target,the inversion method is used to calculate the weights between traces in the pre-stack gather.First,according to the quantitative characteristics of resolution-enhancement seismic data: ’the dominant frequency is high enough and the bandwidth is wide enough’,the repeated weighted stacking tests are performed based on the heuristic optimization algorithm.The global optimization algorithm is applied to obtain the stacking weight of each trace,and then the resolution-enhancement seismic data is obtained.Because of the low computational efficiency of the simulated annealing algorithm and genetic algorithm,it is considered that the stacking weights between continuous gathers should be similar based on the lateral continuity of the seismic profile.The weight obtained by the optimization of the current gather is taken as the initial value of the next gather.This optimization method is similar to the migration problem of the dominant population,named as ’Migration-based Fast Genetic Algorithm(MFGA)’.This algorithm can effectively improve the computational efficiency,and converge to the global optimal solution more easily,thereby improving the seismic stacking resolution.Subsequently,when the dominant frequency is high enough and the bandwidth is wide enough,the seismic amplitude spectrum is approximately constant.According to phase retrieval theory,it is considered that the stacking amplitude spectrum is closest to constant.Based on the above thought,the inversion equation is established,and the nonlinear optimization method of stacking weights is linearized,which greatly improves the computational efficiency and is suitable for the weighted stacking of large-scale three-dimensional data.When seismic attenuation occurs,the amplitude,frequency,and phase are all functions of time.Therefore,we proposed a weighted stacking algorithm using time-window division: the seismic data is divided into multiple time windows,and then the weights are calculated in each time window so that the algorithm is suitable for nonstationary seismic data.Finally,considering the small difference in the amplitude spectrum shape among traces in most seismic gathers,the weighted stacking algorithm cannot obtain the most ideal resolution-enhancement seismic data.Meanwhile,we found that the reconstruction of some time-frequency analysis algorithms is equivalent to seismic stacking.Therefore,we carry out frequency decomposition on seismic gathers and perform signal-to-noise ratio weighted stacking on each frequency component,which is similar to the process of gather optimization.Then we use linear inversion to obtain the optimal stacking weight of each frequency component for secondary stacking.The secondary is the stacking of different frequency components,also known as the reconstruction step of the time-frequency analysis algorithm.By twice weighted stacking algorithm,we can obtain the seismic profile with improved resolution,which avoids the problem of the small difference in the amplitude spectrum shape among traces in most seismic gathers.