论文部分内容阅读
塔河油田4区奥陶系油藏的主要产层为鹰山组碳酸盐岩储层,该区储层经历的地质时期长,受成岩作用、构造运动和岩溶作用改造强烈,形成了不同类型的储集空间,给测井解释带来极大困难。文章针对本地区储层孔隙结构类型多样、储层非均质性严重等难题,将研究区储层分为四种类型:未充填洞穴型、部分(全)充填洞穴型、裂缝—孔洞型和裂缝型,并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,笔者以典型性为原则挑选出自然伽马、深侧向、浅侧向、声波、密度、中子六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。
The main formation of the Ordovician reservoir in Block 4 of Tahe Oilfield is Yingshan Formation carbonate reservoir. The reservoir experienced long geological period and was strongly altered by diagenesis, tectonic movement and karstification to form different Type of reservoir space, logging interpretation to bring great difficulties. In this paper, the reservoir types in the study area are divided into four types according to different types of pore structure and serious reservoir heterogeneity in this area: unfilled cave, partially filled cave, crack-hole and Fracture type, and combined with test mining data qualitative analysis of each reservoir logging response characteristics. On this basis, the author selects six kinds of well logging information of natural gamma ray, deep lateral, shallow lateral, acoustic wave, density and neutron as parameters according to the principle of typicality. According to the disadvantages of conventional BP neural network, Neural network method for automatic classification of reservoir recognition, and achieved good results.