Research On Robust Spectral-modeled Deconvolution

With the opening of China's 14 th Five-Year Plan and the economic restart of countries around the world after the global pandemic of the new crown pneumonia,global development's demand for energy is even greater,which will make it possible to find usable underground oil and gas reservoirs under existing conditions.It's getting more difficult.The development of oil and gas exploration and development has also ushered in new opportunities and challenges.Since most of the oil and natural gas resources are buried deep underground,how to accurately obtain their location distribution information occupies an important position in the process of oil and gas development.Due to the low resolution of seismic data collected in the field,certain processing is required to obtain high-precision and high-quality geological information.How to improve the resolution of seismic data is a long-term research problem of seismic prospectors.The deconvolution method is a commonly used method in the field of seismic data processing to improve the time resolution of seismic data.Compared with other methods of deconvolution,spectral simulation deconvolution has obvious advantages in processing non-white noise seismic data.However,in the process of wavelet fitting,small changes in parameters will lead to larger changes in the waveform of the solution wavelet,causing problems such as unstable and uncontrollable deconvolution results.In order to better improve the resolution of seismic data,we firstly propose a robust spectral simulation deconvolution algorithm based on Lake wavelet based on traditional spectral simulation deconvolution.This method adds the assumption that the shape of the seismic wavelet spectrum is the same as the waveform of the Lake wavelet,thereby reducing the influence of the parameters in the spectrum simulation process,and can robustly solve the wavelet with a known shape.Firstly we use the autocorrelation of seismic data to solve the main frequency of the seismic data wavelet,and then we use a combined wavelet to solve the deconvolution operator,and the operator is applied to the seismic data to intuitively and quantitatively increase the frequency domain amplitude of the data Spectrum,finally the method achieves good and robust spectrum simulation deconvolution.Furthermore,in view of the need to set a greater degree of spectral enhancement when using the robust spectral-modeled deconvolution algorithm based on the ricker wavelet to process data with low signal-to-noise ratio,we firstly use the long short-term memory neural network to extract the non-linear relationship between the different frequency responses of the reflection coefficient sequence,and then the high-frequency information can be predicted by using the information of the seismic data in the medium and low frequencies,so as to improve the high-frequency information of the actual seismic data and achieve the purpose of improving the resolution of the seismic data.Finally,we apply these two methods to improve the resolution of the synthetic data and the actual area seismic data,by comparing and analyzing,and the advantages of the proposed method of improving the resolution of seismic data are verified.