论文部分内容阅读
地震属性数据中包含着一些潜在的有用信息,但从地震资料中提取的地震属性多达数百种,这对实际研究造成很大困难。根据地震属性之间必然存在着相关性这一前提,对地震属性进行降维分析是解决这一困难的有效措施。将数据挖掘技术中的K-L变换和奇异值分解2种线性降维方法,分别应用到地震属性降维中,再将降维后的地震属性应用到研究区的实际地震数据中进行测试,结果发现降维后的地震属性均集中在特征值最大的那个新属性上。并依次减小,认为用降维后的地震属性预测储层比用单一地震属性更符合地质规律。
Seismic attribute data contains some potentially useful information, but there are hundreds of seismic attributes extracted from seismic data, which is very difficult for practical research. According to the premise that there must be correlation between seismic attributes, dimensionality reduction of seismic attributes is an effective measure to solve this problem. The two kinds of linear dimensionality reduction methods of KL transform and singular value decomposition in data mining technology are respectively applied to the dimensionality reduction of seismic attributes, and the seismic attributes after dimensionality reduction are applied to the actual seismic data in the study area for testing. As a result, The seismic attributes after dimensionality reduction are all concentrated on the new attribute with the largest eigenvalue. And decrease in turn. It is considered that using the seismic attributes after dimensionality reduction to predict the reservoir is more in line with the geological laws than using a single seismic attribute.