摘要
塔河油田四区奥陶系油藏主要产层为鹰山组碳酸盐岩储层,该套储层曾长时期受成岩作用、构造运动和岩溶作用的强烈改造而形成不同类型的储集空间,给测井解释带来极大困难.针对该区储层孔隙结构类型多样、储层非均质性严重等情况,将研究区储层分为4种类型:①未充填洞穴型;②部分(全)充填洞穴型;③裂缝-孔洞型;④裂缝型.结合试采资料定性分析了每种储层的测井响应特征.在此基础上,以典型性为原则挑选出自然伽玛、深侧向、浅侧向、声波、密度、中子等6种测井信息作为参数,针对常规BP神经网络的缺点,采用改进BP神经网络方法对储层进行了自动分类识别,取得了较好的效果.
Yingshan Group carbonate reservoir is the dominant Ordovician reservoirs in block 4 of Tahe Oilfield. The Reservoir has undergone strong diagenesis,tectonics and karstification for a long geological periods. Now the different types of reservoir space bring enormously difficulty to the well logging explanation. This paper tackles many difficulties because of the various strong heterogeneous fracture-cave system in this zone. The studied pay zone can be divided into four categories,which are ①unpacked cavern,②partial（full） packed cavern,③fracture-cavern,④fracture, and relevant typical log response are analyzed along with well test. According to principle of typicality, six types of log response information, which consist of GR ray, deep lateral resistivity, shallow lateral resistivity, acoustic velocity, density and neutron porosity. To overcome the shortcomings of conventional BP neural network model, improved BP model is brought forward to realize the classification of the reservoir, and good results have been achieved.
出处
《新疆地质》
CAS
CSCD
2007年第4期405-408,共4页
Xinjiang Geology
关键词
塔河油田
碳酸盐岩储层
神经网络
测井识别
Tahe Oilfield
Carbonate reservoir
neural network
Logging identification
