摘要
在利用多参数进行储层油气预测时,并不是使用的特征越多越好,最佳特征的维数取决于实际问题的预测效果。这里运用聚类分析法优选地震特征参数,将距离较远或相似系数低的特征参数聚为一类,用来对未知样本进行地震储层预测。利用优选后的参数进行神经网络储层油气预测,在实际应用时取得了较好的效果。
It is not always beneficial to use the much more seismic attribute parameters in reservoir prediction.The optimal dimension of the seismic attributes depends upon the effect of reservoir prediction.In this paper,we choose seismic characteristic parameters by the clustering method and by using the optimal attribute parameters,which are of large Euclidean distance and low similarity coefficients to predict reservoir and hydrocarbon,an effective improving in the results is obtained in the practice.
出处
《物探化探计算技术》
CAS
CSCD
2007年第5期420-424,369-370,共5页
Computing Techniques For Geophysical and Geochemical Exploration
基金
上海市重点学科建设资助项目(T0502)
关键词
神经网络
地震特征参数
油气预测
聚类分析
neural networks
seismic characteristic parameters
hydrocarbon prediction
clustering method
