Seismic Attribute Analysis And Reservoir Fluid Recognition In Fractional Domain

Time-frequency analysis technology is an important part of reservoir prediction and interpretation.The time-frequency spectrum of seismic signals can be obtained by time-frequency analysis methods,and then the seismic attributes can be extracted and identified by time-frequency features.Because the time-frequency properties of seismic signals can reflect the inherent characteristics of transmission medium differences,the absorption and attenuation of seismic waves of different properties are not the same.Therefore,the prediction and identification of reservoir fluid by analyzing the time-frequency properties of seismic signals is an effective method.The resolution of short-time Fourier transform is too low,and Wigner-Ville distribution suffer from cross-term interference issue.These defects greatly restrict their applications in seismic exploration.In recent years,the high-precision time-frequency analysis methods have been developed rapidly,such as fractional time-frequency analysis,synchrosqueezing wavelet transform,sparse time-frequency analysis and so on.These methods have been successfully applied to seismic exploration and reservoir prediction,and achieved good results.This dissertation focuses on post-stack seismic data and extract the related seismic attributes based on high-precision time-frequency features to achieve fluid identification.The research contents of this dissertation mainly include the following aspects:(1)A variety of time-frequency analysis methods,such as short-time Fourier transform,Gabor transform,wavelet transform,S transform,Wigner-Ville distribution and Cohen's distribution are studied.The simulation tests of these methods illustrate their respective advantages and disadvantages.(2)The Hilbert-Huang transform is studied,which is an adaptive time-frequency analysis method for nonlinear and nonstationary signals.In this method,there are Empirical Mode Decomposition(EMD)and Hilbert transform these two parts.The principle and implementation process of this method are described in this dissertation.Due to the existence of mode mixing phenomenon in EMD,this dissertation also introduces Ensemble Empirical Mode Decomposition(EEMD),which is an improved algorithm of EMD.The EEMD method can effectively solve the mode mixing problem,but it will lead to some noise in the decomposition results.The calculation of the instantaneous frequency is sensitive to noise,so a more robust method of instantaneous frequency estimation is required.To solve this problem,this dissertation proposes an improved Hilbert-Huang transform based on damped instantaneous frequency.Compared with other frequency estimation methods,the damping instantaneous frequency estimation method has more robustness against noise and higher estimation accuracy.The synthetic model and real seismic data test results show that the method can effectively obtain the instantaneous frequency attributes of seismic signals.(3)The generalized time-bandwidth product,as an extension of the time-bandwidth product in the fractional domain,can measure the support area of the effective signal in the fractional time-frequency domain.Based on the definition of generalized time-bandwidth product and combining with the time-frequency rotation property of fractional Fourier transform,an optimal fractional S transform is proposed,which can achieve higher time-frequency resolution than traditional S transform.In addition,according to the definition of normalized second-order central moment(NSOCM),the optimal order search problem is transformed to the direct calculation of NSOCM,which improves the computational efficiency of the algorithm.The effectiveness of the algorithm is verified through theoretical signal simulation.The spectral decomposition results of real seismic data show that the optimal fractional S transform based on NSOCM can obtain single-frequency visualization with better time-frequency concentration,important in the analysis of hydrocarbon reservoirs.(4)Based on matching pursuit algorithm and sliced Wigner high-order spectra,a matching pursuit-based sliced Wigner higher order spectra is proposed,which can obtain a sparser high-resolution time-frequency spectrum without cross terms.The simulated model and real data test results show that this method can provide single-frequency slices with greater precision,thereby enhancing the precision of reservoir prediction.(5)Based on the inverse problem solving optimization theory,this dissertation proposes a sparse S transform method,which regardes the inverse S transform as a linear inverse problem and adds L1 norm sparsity constraint to the time-frequency spectrum of S transform.In the basis pursuit denoising(BPDN)form,the optimal solution can be obtained by using the SPGL1 algorithm and a high-resolution sparse time-frequency distribution can be obtained.The good performance of the proposed method is assessed on simulated and real seismic data.The results indicate that the sparse S transform can provide a high resolution and focused time-frequency spectrum for seismic data,which is conducive to seismic imaging and reservoir interpretation.(6)In this dissertation,the concept of spectral saliency detection is introduced into the seismic field.Based on the theory of fractional Fourier transform,the seismic saliency detection model in fractional Fourier transform domain is proposed,which is modified during the implementation process to get more smooth and clear seismic saliency characteristics.Taking the fractional spectral residual and fractional phase spectrum two models as examples,their performance and advantages and disadvantages are analyzed by comparing the simulated and real seismic data.By using the saliency detection model in fractional domain,several saliency maps at different fractional orders can be obtained for seismic attribute analysis.These saliency maps can characterize the detailed features and highlight the object areas,which is of great significance in oil and gas reservoir analysis.(7)The time-frequency attribute extraction method of seismic signal is studied,mainly using the optimal fractional S transform to extract a variety of seismic attributes,such as center frequency,root mean square frequency,instantaneous bandwidth and Teager main energy attributes,and compared with the results of conventional S transform.This dissertation also analyzes the low frequency shadow phenomenon in the spectral decomposition results.In addition,the energy attenuation properties are used to identify the fluid present in the seismic data.