【摘要】 将基于PCA-Kmeans++的多属性融合聚类技术应用于沁水盆地南部3#煤层的储层预测中,对融合聚类属性进行分析,确定有利储层分布。首先提取常规的叠后地震属性、叠后波阻抗反演以及叠前AVO属性;然后利用PCA主成分分析方法,得到贡献率最大的几个主成分分量;最后通过Kmeans++无监督机器学习算法对主成分分量进行融合和聚类。实际资料应用结果表明,PCA-Kmeans++方法可以融合各个属性的特征,能够更加清晰地反映地质异常体的分布特征,为沁水盆地南部煤层气及类似储层的预测提供了一种可行的方法。更多还原

【Abstract】 The cluster analysis method for multi-attribute fusion based on PCA-Kmeans ++ was proposed to predict coalbed methane in southern Qinshui basin. Firstly, post-stack seismic attribute,post-stack impandence inversion and pre-stack AVO attribute were obtained. Secondly, principal components of the maximum contribution rate were calculated by principal components analysis(PCA).Finaly, a fusion clustering attribute was generated by Kmean++. The results of actual data show that the proposed method can compromise the characteristic of several attributes and reflect the characteristic of geological anomaly body in this region. The method can be used by domain scientist to predict coalbed methane in southern Qinshui basin and other reservoir similar to Qinshui basin.更多还原