河道砂体含油性预测方法研究

The comprehensive methods to predict oil-bearing capability in channel sand

1. 中国石化胜利油田有限公司地质科学研究院,山东东营 257015;2. 石油大学资源与信息学院,山东东营 257015

Geoligical Science Institute of Shengli Oilfield, SINOPEC, Dongying 257015, China

利用Kohonen网络、模糊神经网络和支持向量机等方法,对胜利油田埕东凸起北坡河道砂体的含油性进 行了预测。对其中1号河道砂体预测结果的分析表明,3种方法各具特点,Kohonen网络法因为使用了多属性聚 类的结果,因此与研究目标的关系比较直观;模糊神经网络法充分考虑了河道砂体内部存在的差异性;支持向量 机法完整地利用了地震波的属性。与井点的含油性对比,3种方法的预测效果中支持向量机法最好,Kohonen 网络法次之,模糊神经网络法稍差。综合应用各种预测方法,可以使预测结果更加准确。

On the base of the analysis of the oil distribution forecast conditions and under the instructions of the geological regularities and characters, the oil distribution forecast methods (Kohonen network, the fuzzy neural network, the support vector machine) of channel sand are bringing forward. Using those methods forecast oil distribution of No. 1 channel sand, and gains corresponding forecast results. Gained corresponding conclusion is as follow, three methods of channel sand oil distribution forecast possess each peculiarity, and forecast results have a little difference. It is consistent for oil distribution forecast posture of No. 1 channel sand. It is efficient and practical to explain oil distribution forecast methods.