神经网络技术在复杂断块油藏水淹层评价中的应用  

Application of neural network technique in evaluating watered- out zone in complex faulted block reservoir

作　　者：张学磊[1] 沈楠[1] 樊茹[1] 张守良 唐玮[1] 

机构地区：[1]中国石油勘探开发研究院,北京100083 [2]中国石油勘探与生产分公司,北京100011

出　　处：《复杂油气藏》2015年第3期55-58,70,共5页Complex Hydrocarbon Reservoirs

基　　金：中国石油重大专项"水驱提高采收率关键技术研究"(2011B-1103)

摘　　要：复杂断块油藏岩性纵向上变化大,水淹后物性和电性也与原始状态有明显差距,使用一般测井解释方法识别水淹层级别具有难度,再加上众多确定的以及不确定的断层的存在,使得从动态上定性识别水淹层也难以做到。以复杂断块G油田试油资料为基础,运用BP神经网络技术,优选对水淹程度敏感的电阻率测井、自然电位测井、声波时差测井以及自然伽马测井数据作为学习数据,建立了水淹层网络训练模型,并据此对未试油小层的水淹层级别进行了预测,证实水淹层评价符合率达到80%以上,由此可以证明BP神经网络技术对此类油田水淹层评价具有很好的适应性。In vertical direction of complex faulted- block reservoir,lithology varies greatly. And then rock physical property between the initial state and watered- out state shows obvious disparity,as well as electrical property. So it is difficult to identify watered- out zones by general log interpretation methods. It is also hard to explain levels of watered- out zone because of the existence of a multitude of definite and undefined faults. In the article,based on formation test data of G Oilfield,resistivity logging,spontaneous potential logging,acoustic logging,gamma ray logging,which are sensitive to watered- out zones,were chosen for learn- training data. And then the BP neural network model for watered- out zones was built. Thus the watered- out zones were predicted by the BP neural network technique. The results proved that the coincidence rate of evaluating watered- out zone is above 80%. Therefore,it was confirmed that the BP neural network technique fits well for evaluating watered- out zone in this type of oilfield.

关 键 词：油田开发 神经网络技术 复杂断块油藏 测井解释 水淹层评价 

分 类 号：TE319[石油与天然气工程—油气田开发工程]