ABSTRACTS

Intelligentseismicfaciesclassificationbasedonanimproveddeeplearningmethod.YANXingyu1,2,GUHanming1,2,LUOHongmei3,YANYouping4.OilGeophysicalProspecting,2020,55(6):1169-1177.

Intelligent seismic facies classification based on deep learning can greatly reduce manual operations.However,when using conventional deep learning methods for seismic facies recognition,the network model can only extract the feature map on a single receptive field,and it is difficult to obtain the global spatial distribution on seismic sections.In addition,the prediction of the boundary of minor seismic facies is inaccurate,and there is not a me-thod for assessing the reliability of prediction on multi-class segmentation models.We propose a facies classification network by simplified U-Net with Pyramid Pooling Module which has been empirically proved to be an effective global contextual prior.And an objective function combining cross-entropy and Dice loss is adopted to improve the boundary characterization of minor seismic facies in unbalanced data.We present Prediction Entropy for estimating the uncertainty of classification results.Applied to F3 dataset,the improved method can enhance prediction accuracy and boundary characterization,and the index of Prediction Entropy can evaluate the uncertainty of the prediction results.

Keywords: seismic facies classification,deep learning,Pyramid Pooling Module,uncertainty estimation

1.Institute of Geophysics & Geomatics,China University of Geosciences,Wuhan,Hubei 430074,China

2.Hubei Key Labrotary for Subsurface Multi-scale Imaging,China University of Geosciences,Wuhan,Hubei 430074,China

3.Research Institute of Exploration & Development,Sinopec Shengli Oilfield Comopany,Dong-ying,Shandong 257015,China

4.North China Branch,Sinopec Engineering Geophysics Co.,Ltd.,Zhengzhou,Henan 450000,China

Characterizationofthinsandreservoirsbasedonamulti-layerperceptrondeepneuralnetwork.DUXin1，2，F(xiàn)ANTing’en1，2，DONGJianhua1，2，NIEYan1，2，F(xiàn)ANHongjun1，2，andGUOBoyang3.OilGeophysicalProspecting,2020,55(6):1178-1187.

Characterization of thin reservoirs is significantly important in seismic exploration.Compared with the method based on seismic inversion,reservoir prediction based on seismic multiple-attribute regression (MAR) can enhance resolution and alleviate the over modeling issue,but the poor ability of gene-ralizing the trained model in MAR frequently causes the instability of the estimated result between wells.A multi-layer perceptron (MLP)-based MAR method is proposed for characterizing thin sandstone-shale reservoirs.This method takes seismic data (background information),90°-phase data (the framework of reservoir structure) and discontinuious reservor boundary (which is a self-deve-loped seismic attribute measuring the discontinuity of reservoir) as inputs,and the gamma-ray (GA) logs of wells as expected outputs,uses the MLP deep neural network to train the model for estimating the GA data,and finally characterizes thin re-servoirs based on the close lithologic relationship between sandstone and shale.Applications to field data from an offshore oilfield A show that the correlated coefficient between the estimated and the true GA of wells has reached 0.855 in a training set with 10 wells,and 0.864 in a prediction set with 2 wells.They are significantly better than the results from the traditional MAR method.Based on the GA data,we interpreted 6 top surfaces for finely describing the reservoir in a target area， and extracted the reservoir-sensitive seismic attribute (Sum of Negative Amplitude,SNA) along the horizon to assess the association between the SNA and the sum thickness (ST) of the reservoir drilled in 156 wells.The SNA based on GA data shows relatively high association with the ST.The correlation coefficient between the SNA based on GA estimation and the ST is about 38% higher than that between the SNA based on the 90°-phase data and the ST,which further confirmed the feasibility of the proposed MLP-based MAR method.

Keywords：thin sand-shale reservoir,multi-layer perceptron,deep learning,seismic multiple-attribute regression

1.State Key Laboratory of Offshore Oil Exploitation,Beijing 100020,China

2.CNOOC Research Institute Ltd.,Beijing 100020,China

3.School of Energy Resources,China University of Geosciences(Beijing),Beijing 100083,China

AutomaticpickingofseismicfirstarrivalsbasedonhybridnetworkU-SegNet.CHENDewu1,YANGWuyang1,WEIXinjian1,LIHaishan1,CHANGDekuan1,andLIDong1.OilGeophysicalProspecting,2020,55(6):1188-1201.

The traditional first arrival picking method cannot take into account picking effect and efficiency,the algorithm stability is poor,and the industrial application has not been very mature.The first arrival picking method based on deep learning is time-consuming and labor-intensive,the process of data preprocessing is cumbersome,and the network structure is too complex,resulting in low training and test efficiency.Combining the advantages of U-Net with those of SegNet,a new hybrid network U-SegNet is constructed,and based on which first arrivals can be picked automatically.Based on the SegNet structure,U-SegNet provides multi-scale information of the encoder network by fusing jump connections information before the deconvolution layer of the decoder network to obtain better performance,and its upsampling operation changes the deconvolution in U-Net to unpooling.Because the pooling index is passed to the upsampling layer,the network model converges faster.Therefore,the U-SegNet network structure is more conducive to segmenting the background noise area and the area where background noise and valid signal overlap,thereby improving the accuracy of first arrival picking.The first arrival automatic picking process based on U-SegNet includes making a training data set,designing a network model,training the network model,testing the network model and applying it to real seismic data.Tests and applications of the U-SegNet model show that the picking efficiency of the proposed method is about 2.2 times that of a commercial software.It is easy to be industrialized and has a good future in large-scale application.

Keywords：seismic first arrival,pick up,deep learning,U-Net,SegNet,U-SegNet

1.Northwest Branch,Research Institute of Petroleum Exploration & Development,PetroChina,Lanzhou,Gansu 730020,China

SuppressingseismicrandomnoisebasedonDeep-KSVD.TANGJie1,MENGTao1,ZHANGWenzheng1,andCHENXueguo2.OilGeophysicalProspecting,2020,55(6):1202-1209.

Curvelet transform denoising causes the events to be distorted and interferes with the effective signals in discontinuous areas such as fault zones.The algorithm of overcomplete dictionaries for sparse representation (K-SVD) requires manual and repeated adjustment of parameters to improve the denoising effect.After comprehensively considering the advantages of deep learning network and sparse representation,we combined the K-SVD denoising algorithm with the deep learning network,and proposed the random noise suppression method based on Deep-KSVD.In order to make the network have the ability to learn parameters,the OMP algorithm is replaced by an equivalent learnable alternative in the tracking phase.The calculation process includes decomposing seismic data into overlapped data blocks,de-noising each data block by proper tracking,and reconstructing the whole data by weighting the denoised data blocks.The denoising process includes three parts: sparse coding,estimation of regularized coefficient and reconstruction of data block.The test results on model data and actual data show that after training a Deep-KSVD network,for given noisy data,it can adaptively attenuate the seismic noises without further adjusting parameters while protecting the effective information of discontinuity and the characteristics of data structure.Compared with the K-SVD denoising method,the Deep-KSVD denoising method provides better effect of noise suppression and can improve the signal-to-noise ratio of full-band data.

Keywords：Deep-KSVD,random noise,suppression,deep learning,sparse signal

1.School of Geosciences,China University of Petroleum (East China),Qingdao,Shandong 266580,China

2.Research Institute of Exploration and Development,SINOPEC Shengli Oilfied Company,Dongying,Shangdong 257105,China

Seismicnoisesuppressionbasedonconvolutionaldenoisingautoencoders.SONGHui1,2，GAOYang3,CHENWei1,4，ZHANGXiang1,2.OilGeophysicalProspecting，2020，55(6)：1210-1219.

Noise attenuation is a long-standing problem in seismic exploration.Traditional denoising me-thods can suppress seismic noise,but they may cause lost effective signals,residual noises and other problems.An unsupervised denoising algorithm based on a convolutional denoising autoencoder is proposed,which can significantly improve the signal-to-noise ratio of seismic data.The algorithm locally and randomly damages seismic data,and then transmits the seismic data damaged to coding and decoding frameworks.The coding framework captures the waveform of the seismic data and eliminates the noise.The decoding framework expands the feature map,recovers the details of the seismic data.Finally,after reconstructing the seismic data,the algorithm trains a model with the error between the reconstructed seismic data and the original seismic data.Considering the complexity of seismic data，a multi-scale convolution module is required to extract the characteristics of seismic data during coding and decoding.Applications to synthetic and real seismic data have proved that the new method is more effective in preserving signals while suppressing noises.Its denoising result is better than a traditional algorithm.

Keywords：unsupervised learning，convolutional neural network，denoising autoencoders，seismic data，denoising

1.Key Laboratory of Exploration Technologies for Oil and Gas Resources(Yangtze University),Ministry of Education,Wuhan,Hubei 430100,China；

2.College of Geophysics and Petroleum Resources,Yangtze University,Wuhan,Hubei 430100,China；

3.CNPC Key Laboratory of Geophysical Prospecting,China University of Petroleum (Beijing),Beijing 102249,China；

4.Hubei Cooperative Innovation Center of Unconventional Oil and Gas,Wuhan,Hubei 430100,China

Theshotinfillingmethodbasedonseismiclegacydata.XUYinpo1，2，SONGQianggong3,PANYingjie1，NIYudong1，ZOUXuefeng4，andYUJianpeng1.OilGeophysicalProspecting,2020,55(6):1220-1230.

Based on ray tracing and wave equation illumination result，the conventional method for improving the imaging quality of the shadow zone of a target layer with complex structures is to analyze the distribution of irradiating energy in the shadow zone and select the infilled shots according to the contribution from illuminated energy to the shadow zone.However，it is impossible to accurately create a geological model due to the complexity of geological structures，which will affect the illumination analysis to complex target layer and lead to it is difficult to improve the imaging quality of the target area with complex structure by infilling shots.Based on this situation，this paper proposes the method for improving the imaging quality of shadow zone of the target layer on the basis of the legacy data according to the local similarity theory：Firstly，the geometry should be drawn up according to the geological task of exploration area，and the shot interval (or shot line interval) should be as small as possible；Secondly，calculate the local similarity of each trace of the common imaging point (CIP) gathers of the legacy seismic data and pick target layer in stacked data；so the local similarity of target layer for all traces of all shots is obtained； Thirdly，use spatial interpolation to calculate the local similarity of all shots and the contribution of the normal shots to the target layer with the designed geometry，and establish the relation curve (curved face) between the local similarity of each CIP gather and spatial position to determine the zones with low local similarity value； Finally， use the uniformity of local similarity of target layer as index to determine the positions and density of infilled shots.The verifications by theoretical data and practical data indicate that this method can economically infill suitable shots and improve the imaging effect of shadow zones of the target layer in areas with high steep dip angle and overthrust structures and whose lateral velocities change severely.

Keywords： geometry，infilled shots，common imaging point (CIP) gathers，local similarity，stacked data，uniformity

1.Acquisition Technique Center，BGP，CNPC，Zhuozhou，Hebei 072751，China

2.Southwest Petroleum University，Sichuan Province Natural Gas Geology Key Laboratory，Chengdu，Sichuan 610500，China

3.Research & Development Center，BGP，CNPC，Zhuozhou，Hebei 072751，China

4.Changqing Division，BGP，CNPC，Xi’an，Shaanxi 710021，China

Depth-domainwaveletestimationusingthesubspace-constrainedHubernorm.ZHANGJie1,2，CHENXuehua1,2，JIANGWei1,2，DANZhiwei3,andXIAOWei3.OilGeophysicalProspecting，2020，55(6)：1231-1236,1244.

Generally,in the constant-velocity depth domain,only a few hundred meters of logging information is available,and is equivalent to 2～5 times the length of a constant depth-domain seismic wavelet.It is difficult to estimate a reliable seismic wavelet from such a short data segment.To address this issue,we proposed a method for estimating the constant-velocity depth-domain wavelet using the subspace-constrained Huber norm.The corresponding procedure can be divided into three steps.First,the reflectivity and the seismic trace near the well location are transformed from the real depth domain to the constant-velocity depth domain.Next,set a threshold and generate the initial synthetic seismogram by convoluting the constant-velocity depth-domain reflectivity and the initial seismic wavelet.Finally,the seismic wavelet is updated by iterative least square method according to the residual error of the synthetic seismogram and the seismic trace until the iteration termination condition is reached.We compare the proposed method and the conventional least square-based methods through the synthetic and field seismic data.The results show that the high performance of our method for estimating the reliable wavelet from limited seismic and well-log data segment.The proposed method can be further extended to estiamte the depth-variant seismic wavelets．

Keywords: depth domain,seismic wavelet estimation,subspace constrained,Huber norm

1.State Key Laboratory of Oil & Gas Reservoir Geology and Exploitation,Chengdu University of Technology,Chengdu,Sichuan 610059,China

2.Key Laboratory of Earth Exploration and Information Technology of Ministry of Education,Chengdu University of Technology,Chengdu,Sichuan 610059,China

3.Data Processing Company,Geophysical Branch,China Oilfield Services Limited,CNOOC,Zhanjiang,Guangdong 524057,China

Adebubblingmethodbasedonextractedwaveletanditscontrollingfactors.MAGuangkai1,ZHOUZhengzheng1,LIJianfeng1,LIPeiming1,ZHANGHong-ying1,andCAIDongdi1.OilGeophysicalProspecting,2020,55(6):1237-1244.

In most cases,far-field wavelets are notional,and they are quite different from acquired wavelets.Therefore,data after debubbling based on far-field wavelets have obvious residual bubbles.This paper proposes a data-driven debubbling method.It extracts wavelets by the cross-correlation of data,calculates the debubbling operator based on extracted wavelets and expected wavelets (without bubbles),and then debubbles data.In addition,the paper analyzes two main factors affecting the debubbling results of the method: ①The length of the debubbling operator.It has an impact on debubbling effect,and it is decided by bubble oscillations; ② White noise.It has a bad influence on the notch of ghosts; and as the white noise increases,it will artificially enhance the notch and reduce the deghosting effect.Therefore,during debubbling,it is better to apply a less white noise.The new method based on extracted wavelets solves the problem of the method based on far-field wavelets.The latter doesn’t consider the influence of acquisition conditions on wavelets.The new method can remove bubble oscillations more effectively,increase signal to noise ratio (SNR) and resolution,and maintain the fidelity of data.All of these have been proven by the processing results of field data.

Keywords: bubble effect,attenuation,wavelet extraction,debubbling operator length,white noise

1.Geophysical Research & Development Center,BGP,CNPC,Zhuozhou,Hebei 072751,China

DeconvolutioninSdomainwithphasecorrectionanditsapplication.ZHOUDonghong1,TANHuihuang1,andWANGWei1.OilGeophysicalProspecting，2020，55(6)：1245-1252.

Inverse Q-filter and time-frequency domain deconvolution can improve the resolution of attenuated traces.However,coefficient gain causes inverse Q-filter very instable,and how to accurately estimate Q value is another key issue.Time-frequency domain deconvolution can well adapt to instable seismic records.However,due to the phase distortion of seismic wavelets during the propagating process,time-frequency domain deconvolution can only remove the influence of wavelet amplitude spectra,but cannot accurately extract wavelet phase spectra,affecting the deconvolution result.We propose to optimize the basic conditions of deconvolution by correcting the phase of attenuated seismic records,and then we can achieve deconvolution in S domain through improved genera-lized S transform.Applications to model and actual seismic data have proved that phase-corrected S-domain deconvolution can get more accurate results,and it is an effective method to identify thin layers and describe stratigraphic pinch-out points.

Keywords: S-domain deconvolution,improved ge-neralized S-transform,phase correction,resolution,pinch-out

1. Bohai Oilfield Research Institute,Tianjin Branch,CNOOC,Tianjin 300452,China

DeblendingofseismicdatabasedonS-transformadaptivefilteringiteration.HUANGDezhi1,HANLiguo1,LIHuifeng2，YANGFeilong2，ZHAOXiaoyu3,andSunNan4.OilGeophysicalProspecting,2020,55(6)：1253-1262.

S-transform is a time-frequency analysis method developed from wavelet transform and short-time Fourier transform.After NMO correction,the amplitude and phase of the signal in every seismic trace of a CMP gather are consistent at the same time,and the blended noises in blended seismic data are distributed randomly.Therefore,the distribution of noises and signals in the S-transform spectra of each trace can be effectively determined by stacking CMP traces after NMO correction,and using the S-transform spectra of the stacked trace as references,the designed filter of S-transform spectra can separate the blended noises.In this study，we first designed a filter based on the deviation between the S-transform spectra of seismic data in CMP gathers and the S-transform spectra of stacked traces after NMO correction,and then extracted signals and deblended noises after multi-level adaptive filtering and multiple iterations.Applications to theoretical data and simulated actual seismic data have proved the method can effectively extract signals and separate deblended noises and random noises.

Keywords: deblending,S-transform,adaptive filter,blending noise

1.College of Geo-exploration Science and Technology,Jilin University,Changchun,Jilin 130026,China

2.School of Earth Sciences and Engineering,Xi’an Shiyou University,Xi’an,Shaanxi 710065,China

3.North China Branch,Sinopec Geophysical Corporation,Zhengzhou,Henan 450000,China

4.Research Institute of Exploration and Development,Northeast Oil & Gas Branch of Sinopec,Changchun,Jilin 130062,China

ParabolicRadontransformparallelalgorithmforCPU-GPUheterogeneousplatform.ZHANGQuan1，2，LINBaiyue1，YANGBo3,PENGBo1，ZHANGWei1，andTURan1.OilGeophysicalProspecting，2020，55(6)：1263-1270.

Parabolic Radon transform is widely used for suppressing and removing multiple waves in pre-stack seismic data.Although parabolic Radon transform in mixed domain can well suppress multiples，it still needs a long time when processing seismic gathers larger and larger.In this paper， first we use the GPU to optimize the parabolic Radon transform algorithm in parallel，and use the CUDA library and other optimizing techniques to accelerate the most time-consuming Fourier transform and algebraic operations to nearly 13x speedup over serial execution in the calculating process；then based on the CPU-GPU heterogeneous platform，we propose a CPU-GPU parallel scheme，which makes full use of computer hardware，to realize the parabolic Radon transform parallel algorithm through CPU multi-threading and multiple GPUs in parallel.The speedup can be nearly 30x over serial execution.

Keywords：parabolic Radon transform，multiple suppression，GPU parallel computing，heterogeneous computing

1.School of Computer Science，Southwest Petroleum University，Chengdu， Sichuan 610500，China

2.School of Information and Communication Engineering，University of Electronic Science and Technology of China，Chengdu，Sichuan 611731， China

3.State Grid Chongqing Electric Power Research Institute，Chongqing，404100，China

ApplicationofCFS-PMLincomplexvelocitymodels.YANGXina1，andZHANGWei2,3.OilGeophy-sicalProspecting，2020，55(6)：1271-1281.

In numerical simulation of seismic wave，the limited space size of discrete grids usually leads to artificial reflections when calculating the outermost layer of the model.The perfectly matched layer (PML) method has achieved good results in absorbing artificial reflections，of which the complex frequency-shifted perfectly matched layer (CFS-PML) is particularly effective.Many studies that discussed the optimal settings of the CFS-PML parameters only consider homogeneous velocity models.However，there are usually complex velocity models with strong heterogeneous in field applications.Therefore，the CFS-PML parameters setting method of complex velocity models is explored based on the setting method of homogeneous velocity models.It is considered that the parameter setting is not only related to the source and grid properties，but also to the velocity of the absorption layer and the wavelength，that is，highly dependent on the velocity model.For a complex velocity model，the key to parameter setting is how to select a reference velocity，and three practical selection methods of reference velocity are given.By testing and comparing the absorption effects of three complex models，i.e.double-layer，multi-layer and Marmousi models， the best reference velocity selection method is defined as the median velocity of boundary medium，and then the setting method of CFS-PML optimization parameters for complex velocity models is proposed.

Keywords： numerical simulation，complex frequency-shifted perfectly matched layer(CFS-PML)，parameter settings，complex velocity model

1.School of Earth and Space Sciences，University of Science and Technology of China，Hefei，Anhui 230026，China

2.Department of Earth and Space Sciences，Sou-thern University of Science and Technology，Shen-zhen，Guangdong 518055，China

3.Shenzhen Key Laboratory of Deep offshore Oil and Gas Exploration Technology，Southern University of Science and Technology，Shenzhen，Guangdong 518055，China

Seismicwavesimulationusingatrapezoidgridpseudo-spectralmethod.TANWenzhuo1，WUBangyu1，LIBo2，andLEIJun3.OilGeophysicalProspecting，2020，55(6)：1282-1291.

The numerical solution to wave equation is the computational engine of many high-precision imaging and inversion methods in seismic exploration.In order to ensure the accuracy，regular methods generally set a fixed spatial sampling interval according to the minimum velocity of the model.This always causes over sampling in high velocity layers,and the calculation is redundant to some extent.Under the compaction of rocks cause by gravity，the wave propagation speed usually increases along with depth.The trapezoid coordinate transformation can incorporate the general increasing trend of wave velocity.So，we propose seismic wave simulation using a trapezoid grid pseudo-spectral method.We use the trapezoid grid based on trapezoid coordinate transformation to divide medium.The fine grid is used in the shallow low velocity region and the coarse grid is used in the deep high velocity region，which can effectively reduce the number of sampling points.Meanwhile，considering the accuracy of wave field simulation，we use pseudo-spectral method to solve the acoustic wave equation with variable coefficients after coordinate transformation and use the perfectly matched layer to eliminate the fictitious return wave caused by artificial boundary.The test on Marmousi model shows that comparing with the general method，the number of grids generated by trapezoidal grid can reduce 69%，and comparing with the regular grid pseudo-spectral method and the high order finite difference method，the calculation time of trapezoidal grid pseudo-spectral method can respectively reduce 58% and 60%.Therefore，this method is an efficient and high-precision seismic wave simulation method.

Keywords：trapezoid coordinate transform，seismic wave simulation，pseudo-spectral method，acoustic equation

1.School of Mathematics and Statistics，Xi’an Jiaotong University， Xi’an，Shaanxi 710049，China

2.Sinopec INOPEC Geophysical Research Institute，Nanjing，Jiangsu 211103，China

3.The Sixth Gas Production Plant，Changqing Oilfield Company，PetroChina，Xi’an，Shaanxi 710018，China

Impactofshaleanisotropyonthemicroseismicwavefieldofpassivesources.WURuyue1,2,3,LIHan1,2,CHANGXu1,2,YANHongyong1,2,WANGYibo1,2.OilGeophysicalProspecting,2020，55(6)：1292-1304.

Shale reservoir is featured by strong anisotropy. Ignoring the impact of shale anisotropy on the microseismic wavefield of passive sources (induced by fracturing stimulation) could affect the credibility of reservoir characterization. We proposes a method for quantitatively evaluating the deviation in travel time and amplitude of seismic wavefields,and study the influence of shale anisotropy on the microseismic wavefield of passive sources based on synthetic numerical test. Firstly,we use moment tensor to define explosive and shear microseismic sources,and caculate the microseismic wavefield of passive sources based on the elastic wave equation for three-dimensional anisotropic media and the staggered-grid finite-difference algorithm. Secondly,by comparing theoretical wavefields induced by different sources in different anisotropy media,we calculate the deviation in travel time and amplitude of P wave and S wave relative to isotropic models,and quantitatively and qualitatively analyze the impact of shale anisotropy on the microseismic wavefields under different source mechanisms. Numerical test show that for a surface microseismic monitoring system and a horizontally layered VTI medium model,shale anisotropy has significant influence on the travel time and amplitude. The travel time can deviate by 30%,and the deviation of amplitude can be up to 60%. The deviation of the travel time increases with increasing offset. In addition,under different source mechanisms,the impact of shale anisotropy on the microseismic wavefield are similar to each other.

Keywords: microseismic,shale anisotropy,source mechanism, direct wave, amplitude, travel time, numerical test

1.Key Laboratory of Petroleum Resource Research,Institute of Geology and Geophysics,Chinese Academy of Sciences, Beijing 100029

2.Institution of Earth Science, Chinese Academy of Sciences, Beijing 100029

3.University of Chinese Academy of Sciences,Beijing 100049

Least-squaresreverse-timemigrationbasedonreflectiontheory.DUANXinbiao1，2,WANGHuazhong1,andDENGGuangxiao3.OilGeophysicalProspecting，2020，55(6)：1305-1311.

Least-squares migration can provide an accurate solution to the model space based on the linear inversion theory.Compared to conventional migration methods，least-squares migration can increase imaging resolution，improve amplitude preservation and reduce migration artifacts.The classical least-squares migration is based on the linearized expression of the scattered wave，and its goal is essentially estimating the scattering intensity of underground medium.While real underground medium is mainly layered，the target of least-squares migration in production is reflection-coefficient imaging.This paper focuses on the difference between linearized expression method of scattering theory and reflection theory.According to the requirement of migration imaging in actual production，then the demigration method based on linearized expression of reflection theory and the least-squares RTM process for estimating the reflection coefficient are established.Sigsbee2a model test shows a good imaging effect using the proposed method.An application of three-dimensional field data shows that the least-squares RTM method is better than RTM.The string beads imaging has better convergence effect，and the deep large faults and small interbed faults are more clearly described.

Keywords：least-squares migration，scattering theory，reflection theory，reflection coefficient，string beads imaging

1.Wave Phenomena and Intelligent Inversion Imaging Group (WPI)，School of Ocean and Earth Sciences，Tongji University Shanghai 200092，China

2.Sinopec Geophysical Research Institute，Nanjing，Jiangsu 211103，China

3.Sinopec Northwest Company,Urumqi,Xiujiang 830011，China

Stableviscoacousticreversetimemigrationinfrequencydomainforundulatedshallowsurface.LIUYanli1,LIZhenchun1,WANGJiao1,SUNMiaomiao1,andLIUQiang1.OilGeophysicalProspecting,2020,55(6):1312-1320.

The energy of seismic wave decreases exponentially as it propagates underground.This results in the analytical solution of conventional viscoacoustic reverse time migration (Q-RTM) increasing exponentially with frequency,and consequently instable wave field compensation,and even failed migration due to the effective wave field covered by high-frequency noise.Based on available knowledge,a stableQ-RTM method in frequency domain for undulated shallow surface is proposed.To restrain the growth of high-frequency components,a stable factor is added to the frequency domain viscoacoustic equation,and its corresponding stable expression forQ-RTM is derived.The stable factor only changes the computing sign of the high frequency outside the effective frequency band,but not bringing additional calculation.The specific process is as follows: ①Establish body-fitted grids and mapping the relationship between physical space and computational space; ②In the computational space,calculate the forward and backward wave fields withQin frequency domain and obtain corresponding wave fields in time domain through IFFT; ③Obtain the forward and backward wave fields in physical space by the mapping relationship; ④Migrate according to the cross-correlation imaging conditions for every time step.Models and raw data have demonstrated that the proposed method can compensate amplitude and high frequencies caused by attenuation,and significantly improve seismic resolution and imaging quality of shallow surface.

Keywords: undulated shallow surface,frequency domain,viscoacoustic reverse time migration,stable factor,body-fitted grid

1.School of Geosciences,China University of Petroleum (East China),Qingdao,Shandong 266580,China

Elasticfullwaveforminversionbasedonfrequency-divisionencoding.SHAOXiangqi1,2,HEBingshou1,2,andSHICaiwang3.OilGeophysicalProspecting,2020,55(6):1321-1329.

The calculation amount and efficiency limit the application of full waveform inversion technology.Source coding technology can effectively reduce the calculation amount of full waveform inversion and improve the inversion efficiency.The traditional source coding technology requires that each gun has the same receiver array,and it has some shortcomings in the adaptability of the recording system,so it can not be directly applied to the recording system with rolling array.In this paper,an elastic full wave inversion method based on frequency-division coding is proposed.Its basic idea is: In forward process,multiple shot records can be extended at the same time,but each shot record is assigned with a harmonic source of different frequency,so the wave fields forward extended at the same time do not coincide with each other in the frequency spectrum.In the process of forward and residual backward propagation of the shot point wave field,phase sensitivity detection is used to separate the single-frequency wave field of each gun to realize the complete separation of the wave fields of each gun,and then the waveform inversion gradient is constructed to update the model.Compared with the traditional source encoding method,the simultaneous interpreting method can not only be speeded up in inversion,but also be applied to the recording system with rolling array in a wider field.

Keywords: elastic wave,full waveform inversion,source code,frequency division code,phase sensitivity detection

1.Key Laboratory of Submarine Geoscience and Propecting Technology,Ministry of Education,Ocean University of China,Ocean University of China,Qingdao,Shandong 266100,China

2.Laboratory of Marine Mineral Resources Evaluation and Detection,Qingdao,Shandong 266100,China

3.Department of Earth and Space Science,South University of Science and Technology,Shenzhen,Guangdong 518000,China

PredictionofTOCcontentinsourcerocksinsouthwesternBozhongSag.WANGXiang1,2,3,MAJinfeng1,2,3,WANGDeying4,WANGZhenliang1,2,3,ZHANGXintao4,andWANGFeilong4.OilGeophy-sicalProspecting,2020,55(6):1330-1342

The total organic carbon (TOC) content in source rocks is the primary parameter for evaluating the abundance of organic matter.Core measurement by geochemical methods can only obtain discrete TOC,and it is difficult to achieve three-dimensional quantitative evaluation on source rocks.In this study,combining geophysical logging data with seismic pre-stack inversion,and aiming at the deep continental source rocks in the southwestern Bozhong Sag,we analyzed and optimized the elastic parameters sensitive to TOC through petrophysical intersection diagrams and established a seismic inversion model of TOC and P wave velocity and density.Using the optimized extended ΔlogR me-thod and seismic pre-stack inversion method,we calculated the TOC of the Dongying and Shahejie formations in the southwestern part of the Bozhong Sag.Finally,we obtained a TOC curve and a 3D TOC inversion body,and calculated the TOC seismic prediction plans of the E3d2L,E3d3,E2s1+2and E2s3members.The results show that the well-seismic prediction error is small,proving the reliability of the prediction me-thod.The source rock in the study area is highly hetero-geneous,and characterized by “multiple layers and wide distribution”.The source rocks in the northern Dong-ying formation are better than those in the southern,and the TOC of the E3d3is generally higher than the E3d2L.The Shahejie formation has higher TOC,and it is the primary source rock interval in the study area,showing gradually decreasing TOC outward from the depositional centers,especially in Wells A,B,C and E drilled into the E2s1+2and their nearby areas.In Wells A and C drilled into the E2s3,and their west and south,the TOC is the highest,showing a good hydrocarbon generating potential.

Keywords: source rocks,TOC,Bozhong Sag,prestack inversion,petrophysics,sedimentary facies

1.Department of Geology,Northwest University,Xi’ an,Shaanxi 710069,China

2.National & Local Joint Engineering Research Center of Carbon Capture and Storage Technology,Department of Geology,Northwest University,Xi’ an,Shaanxi 710069,China

3.State Key Laboratory of Continental Dynamics,Northwest University,Xi’an,Shaanxi 710069,China

4.Exploration and Development Research Institute,Tianjin Branch of CNOOC,Tianjin 300452,China

Pre-stackspectrumblueingfrequencyincreasingtechnique:AcasestudyonreservoirpredictioninChadBaobOilfield.LIXianbing1,ZHAOJunjie2,JINJianli1，LIXiangling1，LIUHanyang3,andJIANGLanting2.OilGeophysicalProspecting,2020,55(6):1343-1348.

In Chad Baob oilfield, the reservoirs are very heterogeneous,and the sand bodies change rapidly.Seismic data with low resolution and poor well-to-seismic correlation make it much difficult to identify the thin interlayers and delineate the lithology boundary.To solve the problem,the pre-stack spectrum blueing frequency increasing technique is applied to processing procedure.Taking the spectrum trends of the well logs and gathers’ forward model as the target,the high frequency energy with serious attenuation in prestack data is recovered,and the resolution of seismic data is improved to the greatest extent within the effective frequency band range.At the same time,the well-to-seismic correlation degree is used to quantitatively control the processing results.The resolution of seismic data, well-to-seismic correlation and the recognition ability of the thin layer are improved by using the method.The application in Chad Baob oilfield shows that the pre-stack spectrum blueing frequency increasing technique provides seismic data with high resolution and high reliability for thin interlayer recognition and reservoir characterization.

Keywords：pre-stack spectrum blueing, pre-stack gather optimization, quantitative quality control of well-to-seismic calibration,seismic inversion

1.Research Institute of Petroleum Exploration & Development,PetroChina,Beijing 100083，China

2.Beijing Energy Star Technology Co.,Ltd.,Beijing 100085,China

3.QITeam (Beijing) Co.,Ltd.,Beijing 100012,China

Faultidentificationbasedonadiscontinuousanisotropicdiffusionfilter.WANGJing1,ZHANGJunhua1,FENGDeyong2,andLIHongmei2.OilGeophysicalProspecting,2020,55(6):1349-1357.

It is of great importance for subsequent seismic interpretation to improve the signal-to-noise ratio of seismic data while preserving the boundary information of faults and other geological bodies.Based on previous studies,we proposed an anisotropic diffusion filter based on discontinuity information.After analyzing the principle of anisotropic diffusion filtering and discussing the relationship between eigenvalues of the structural tensor and local structural features of 3D seismic images,we redesigned the eigenvalues of the diffusion tensor using fault confidence parameters,which can control the filtering intensity to seismic data in different directions.The value of fault confidence is close to zero where there are flat continuous reflectors and the intensity of diffusivity is strong.On the contrary,the diffusion is very weak within presumptive fault zones.Applications to synthetic model and real data have proved that this method can effectively suppress noises,preserve faults and enhance the continuity of reflectors.They are basic data for seismic interpretation.

Keywords: anisotropic diffusion filter,structure tensor,fault confidence,diffusion tensor,eigenvalue

1.School of Geosciences,China University of Petroleum (East China),Qingdao,Shandong 266580,China

2.Geophysical Research Institute,Shengli Oilfield Branch Company,SINOPEC,Dongying,Shandong 257022,China

ShiftsamplingDFT(SFT)theoryanditsapplicationingravityandmagneticprospecting.CHAIYupu1,andWANHaizhen1.OilGeophysicalProspecting,2020,55(6):1358-1363.

The procedure that the Gaussian quadrature rule is introduced into Fourier transformation calculation is shown in rigorous mathematical deduction based on the Shift sampling DFT (SFT) theory.The conclusion is that a Fourier integral can be approached with high precision by the weighted sum of a series of SFT with different shift parameters.The weigh is a half of the Gaussian quadrature coefficient,and the sampling parameter is a half of the Gaussian quadrature node plus 0.5.This conclusion provides a rigorous theoretical basis for improving the precision of forward modeling of potential fields (gravity and magnetic) in wave-number domain.Since the sufficient conditions of both the SFT theory and the Gaussian quadrature theory are limited real functions,the Gaussian FFT algorithm based on the conclusion are applicable in forward and inverse Fourier transforms of any limited real functions.

KeywordsShift sampling DFT theory,Gaussian quadrature theory,forward modeling of potential fields in wave-number domain

1.GME & Geochemical Surveys,BGP,CNPC,Zhuozhou,Hebei 072751,China

Researchondirectintegrationalgorithmofelectromagneticfieldinhomogeneouslayeredmediaforawiderangeoffrequenciesandtransceiverdistances.DAIShikun1,2,ZENGLing1,2,ZHOUYinming1,2,3,LIKun1,2,CHENQingrui1,2,andLINGJiaxuan1,2.OilGeophysicalProspecting，2020,55(6):1364-1372,1394.

In geophysical exploration,the analytic expression of electromagnetic field in homogeneous layered media is Hankel integral,whose kernel function is Bessel function of order 0 and 1.With the increase of its synthesis,Bessel function pre-sents the characteristics of fast oscillation and slow attenuation,which makes Hankel integral difficult to calculate with high efficiency and accuracy,especially in the case of high frequency and large receiving distance.In order to solve this problem,this paper proposes a direct integration method which is efficient and of high precision.The basic idea is that Bessel function can be divided into two intervals and expanded by different polynomials.Hankel integral of each interval can be divided into the sum of multiple unit integrals.The integrated function of each unit is represented by a cubic spline interpolation function,from which the analytic solution of the integral can be obtained.The numerical solution of Hankel integral can be obtained by superposition.On this basis,using the analytical solution to the electromagnetic field of the electric dipole in the uniform whole space,the correct selection of the integral range and the appropriate partition of the integral element are studied.The comparison between the numerical solution and the a-nalytical solution shows that the algorithm is correct and reliable.The comparison between the algorithm proposed in this paper and the digital filter-ing algorithm shows that the algorithm is widely applicable to the calculation of electromagnetic fields with different frequencies and different receiving distances.Therefore it is used very universally.

Keywords: Hankel integral,direct integration method,high frequency electromagnetic field,layered media

1.School of Geosciences and Info-physciences,Central South University,Changsha,Hunan 410083,China

2.Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring of Ministry of Education,Central South University,Changsha,Hunan 410083,China

3.GME & Geochemical Surveys,BGP,CNPC,Zhuozhou,Hebei 072751,China

Multi-referencearrayMTdataprocessingmethod.ZHOUCong1,2,TANGJingtian3,4,YUANYuan1,DENGJuzhi1,SHIFusheng1,andLIYong2.OilGeo-physicalProspecting,2020,55(6):1373-1382.

Rational use of multiple remote reference station data is an important way to improve the quality of magnetotelluric data processing.Based on a processing model of electromagnetic array,a data processing algorithm based on multiple reference stations is proposed in this paper.Firstly,select multiple remote reference stations from both inside and outside the measured area and construct a re-ference data array dominated by natural fields,then extract the polarization parameters of natural sources from the reference data array by robust principal component analysis; and finally apply the parameters to the target array,which is constructed by data from the stations to deal with,and estimate the target tensor impedance through robust estimation.A case study on real data with MT “dead band” distortions indicates that the algorithm proposed in this paper is more effective than conventional methods.The results show that the processing method based on multiple remote reference array stations can improve the quality of MT data processing in a strongly noisy environment and provide a feasible solution to correct MT data with “dead band” distortions.

Keywords: electromagnetic prospecting,magnetotellurics,remote reference,multi-reference array,MT data processing

1.State Key Laboratory of Nuclear Resources and Environment,East China University of Technology,Nanchang,Jiangxi 330013,China;

2.Key Laboratory of Geophysical Electromagnetic Probing Technologies,Ministry of Natural Resources,Langfang,Hebei 065000,China;

3.Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring(Central South University),Ministry of Education,Changsha,Hunan 410083,China;

4.Technical Innovation Center of Coverage Area Deep Resources Exploration,Ministry of Natural Resources，Hefei,Anhui 230001,China

Areviewofnoisesuppressionmethodsformarinecontrolledsourceelectromagneticsignals.LISuyi1，ZHANGYi1,ZHANGJikun1,andSHENChun2.OilGeophysicalProspecting,2020,55(6):1383-1394.

The marine controlled source electromagnetic (MCSEM) method is an emerging marine geophy-sical exploration method for detecting mineral resources such as submarine oil and natural gas re-servoirs.In the complex marine environment,signals are very susceptible to interferences from various noises,which affects data inversion and interpretation.This paper first introduces the principle of MCSEM and key data preprocessing methods,then summarizes the methods for suppressing random noise,‘a(chǎn)ir wave’ noise and seafloor currents,and finally analyzes the future direction on how to denoise MCSEM data.

Keywords：marine controlled source electromagnetic method,random noise,air wave noise,the noise of seafloor currents,noise suppression

1.College of Instrumentation and Electrical Engineering,Jilin University,Changchun,Jilin 130061,China

2.College of Computer Science and Technology,Jilin University,Changchun,Jilin 130012,China

Severalurgentproblemsfacedbymulti-componentseismicsinoilandgasexploration.WANGYun1,WENPengfei2,LIZongjie3,LIUJia1,LIMengze4.OilGeophysicalProspecting,2020,55(6):1395-1406.

We first introduce present development of seismic anisotropy theories and multi-component seismic technology,and then aiming at the interests of petroleum industry,discuss the problems faced by the equivalent medium theories and their applications at home and aboard.It is obvious that the limitation of Thomsen’s weak anisotropic theory needs to be re-considered for strong anisotropic media such as shale reservoirs and coal seams.To describe multi-azimuth and multi-scale fractured reservoirs,we need to start our research on monoclinic medium and improve the applicable equivalent theories.And for continental sedimentary reservoirs widely distributed,especially thin layered formations in China,we need to set up an orthogonal anisotropic model and re-delineate the characteristics of wave propagation to discriminate what causes P-wave anisotropic responses and what links shear-wave splitting and PP’s anisotropics with respect to fractured reservoirs.Also we analyze the defects ofX- andY-component processing and multi-wave interpretation，and conclude that，the processing technique of multi-component seismic data can’t match the need of oil and gas exploration and production,which seriously obstacles the application of multi-component seismics.Finally,we propose the most important researches to be started and traced,including methods of how to keep vector and kinetic characteristics during multi-component processing,methods of simultaneous and high-dimensional interpolation ofX，Y，Zcomponents,softwares for OVT sorting and stacking feasible to wide-azimuth PS data，methods of migrating PS-wave in OBS or OBN data with greatly rising and falling seafloor.

Keywords：seismic anisotropy,multi-component,fracture,equivalent medium theory,vector,kinetic characteristic

1.MWMC Group,School of Geophysics and Information Technology,China University of Geosciences(Beijing),Beijing 100083,China

2.Key Laboratory of Marine Mineral Resources,Ministry of Natural Resources,Guangzhou,Guangdong 510075,China

3.Research Institute of Petroleum Exploration and Production,Sinopec Northwest China Oilfield Company,Urumuqi,Xinjiang 830011,China

4.BGP International,CNPC,Zhuozhou,Hebei 072751,China