随着复杂储层地震资料特征筛选的机器学习技术的进步,如何有效地对参与地震属性优选和储层反演的地震样本进行采集和分析,成为目前智能地震预测领域的一个研究热点。目前的方法多着重于模型分类算法的改进,在标签的制作和采集方面不仅...随着复杂储层地震资料特征筛选的机器学习技术的进步,如何有效地对参与地震属性优选和储层反演的地震样本进行采集和分析,成为目前智能地震预测领域的一个研究热点。目前的方法多着重于模型分类算法的改进,在标签的制作和采集方面不仅耗费大量时间进行人工标注,还存在标签不平衡情况下类内可靠性、类间平衡性不强等问题。为此,提出基于稀疏强特征提取的三维地震数据完备方法。首先,基于多数决原则的样本分割(Sample Segmentation Based on Majority Rule,SSMR)寻迹多尺度、多标签三维地震样本,进行采集、自动标注;然后,改进标签洗牌平衡方法(Improved Label Shuffling Balance Method,ILSB),通过“2+1”的样本增广平衡策略进行数据完备处理,改善样本采样不平衡性导致的模型训练偏向性;最后,利用基于最小L_(1)范数稀疏表示对奇异值分解结果进行强特征提取(Minimum L_(1)-norm Based Sparse Representation for Feature Extraction,L_(1)-SRFE)和可视化表示。实际资料应用表明,实钻井与验证井预测结果吻合度高,该方法具有较高的标签分类准确率。展开更多
Shallow surface wave methods are mostly used for investigation of the surface velocity structure in environmental and engineering geophysics in non-desert areas. For the special geological features of the Takelamagan ...Shallow surface wave methods are mostly used for investigation of the surface velocity structure in environmental and engineering geophysics in non-desert areas. For the special geological features of the Takelamagan Desert area, we use the multi-channel analysis of surface wave （MASW） method to process multi-channel shallow surface wave records to determine the near surface velocity structure in the desert area. We also process, analyze, and compare the surface waves in many-trace records extracted from the oil exploration shot gathers in the area. We show that the MASW method can determine detailed shallow velocity structure in desert areas and the many-trace records can be used to get detailed deep geological structure. The combination of the two different datasets can obtain the exact velocity structure upper 60 m depth in the survey area.展开更多
Conventional deconvolution methods improve seismic resolution at the cost of reduced signal-tonoise ratio(SNR),limiting the accuracy of high-frequency signal recovery.To address this issue,this paper proposes a high-r...Conventional deconvolution methods improve seismic resolution at the cost of reduced signal-tonoise ratio(SNR),limiting the accuracy of high-frequency signal recovery.To address this issue,this paper proposes a high-resolution processing method based on low-dimensional manifold constraints.First,datadriven manifold learning is employed to construct neighborhood relationships and characterize the distribution of high-dimensional seismic records in low-dimensional manifold space.Then,manifold information is incorporated into the regularization framework of high-resolution inversion to establish a multi-channel inversion objective function with low-dimensional manifold constraints.Finally,an iterative optimization strategy is applied for simultaneous multi-channel inversion of reflection coefcient sequences.By introducing spatial correlation of seismic signals into the high-resolution processing workflow,this method effectively suppresses noise interference in high-frequency signal recovery.Both synthetic and eld data tests demonstrate that the proposed method maintains superior SNR while enhancing resolution,improving the characterization accuracy of thin-layer hydrocarbon reservoirs.展开更多
文摘随着复杂储层地震资料特征筛选的机器学习技术的进步,如何有效地对参与地震属性优选和储层反演的地震样本进行采集和分析,成为目前智能地震预测领域的一个研究热点。目前的方法多着重于模型分类算法的改进,在标签的制作和采集方面不仅耗费大量时间进行人工标注,还存在标签不平衡情况下类内可靠性、类间平衡性不强等问题。为此,提出基于稀疏强特征提取的三维地震数据完备方法。首先,基于多数决原则的样本分割(Sample Segmentation Based on Majority Rule,SSMR)寻迹多尺度、多标签三维地震样本,进行采集、自动标注;然后,改进标签洗牌平衡方法(Improved Label Shuffling Balance Method,ILSB),通过“2+1”的样本增广平衡策略进行数据完备处理,改善样本采样不平衡性导致的模型训练偏向性;最后,利用基于最小L_(1)范数稀疏表示对奇异值分解结果进行强特征提取(Minimum L_(1)-norm Based Sparse Representation for Feature Extraction,L_(1)-SRFE)和可视化表示。实际资料应用表明,实钻井与验证井预测结果吻合度高,该方法具有较高的标签分类准确率。
文摘Shallow surface wave methods are mostly used for investigation of the surface velocity structure in environmental and engineering geophysics in non-desert areas. For the special geological features of the Takelamagan Desert area, we use the multi-channel analysis of surface wave （MASW） method to process multi-channel shallow surface wave records to determine the near surface velocity structure in the desert area. We also process, analyze, and compare the surface waves in many-trace records extracted from the oil exploration shot gathers in the area. We show that the MASW method can determine detailed shallow velocity structure in desert areas and the many-trace records can be used to get detailed deep geological structure. The combination of the two different datasets can obtain the exact velocity structure upper 60 m depth in the survey area.
基金supported in part by the Fundamental Research Project of China National Petroleum Corporation(CNPC)under Grant 2022DQ0604-4。
文摘Conventional deconvolution methods improve seismic resolution at the cost of reduced signal-tonoise ratio(SNR),limiting the accuracy of high-frequency signal recovery.To address this issue,this paper proposes a high-resolution processing method based on low-dimensional manifold constraints.First,datadriven manifold learning is employed to construct neighborhood relationships and characterize the distribution of high-dimensional seismic records in low-dimensional manifold space.Then,manifold information is incorporated into the regularization framework of high-resolution inversion to establish a multi-channel inversion objective function with low-dimensional manifold constraints.Finally,an iterative optimization strategy is applied for simultaneous multi-channel inversion of reflection coefcient sequences.By introducing spatial correlation of seismic signals into the high-resolution processing workflow,this method effectively suppresses noise interference in high-frequency signal recovery.Both synthetic and eld data tests demonstrate that the proposed method maintains superior SNR while enhancing resolution,improving the characterization accuracy of thin-layer hydrocarbon reservoirs.
