多信息融合的页岩油储层自动分层技术

Automatic layering of shale oil reservoir with multiple information

页岩油属于非常规资源,其储层"甜点"是主要的勘探目标,有必要将泥页岩层系进行细分。而页岩油储层具有薄互层多、岩石物理差异不明显的特点,单一的测井曲线并不能全面地反映地层信息,给利用测井曲线自动分层及开展后续工作带来了难度。通过挑选纵向分辨率高、对分层敏感的测井曲线,采用滑动加权滤波去噪处理,在消除随机干扰带来的噪声的同时,保护薄层信息不被平滑掉;利用分类主成分分析降维处理,在消除相关性(重叠信息)的同时,降低多重相关性对降维结果的危害。整合多条测井曲线上的薄层信息,形成一条综合了多条测井曲线信息的分层曲线,既有效利用了多条曲线信息,又提高了信噪比,有效地保护了薄层信息。利用可操作性强、原理简单的活度分层法和拐点法验证该技术,均取得了良好的效果,可自动划分出大部分薄层,证实了该方法对薄互层等复杂地层具有一定的有效性和可行性。

＂Sweet points＂ are main exploration targets for shale oil reservoir. It is necessary to subdivide the reservoir into layers. Shale oil reservoir is characterized by multiple thin interbeds and unapparent petrophysical features. A single log can not accurately reflect its comprehensive formation, which brings challenges to automatic layering based on logs and subsequent activities. This paper proposes a new layering method with multiple information. Firstly, logs with high vertical resolution and high sensitivity to layer boundaries are selected, and the selected logs are processed by the sliding weighted filtering method to remove random noises without compromising thin-layer information. Then, the filtered logs are dimensionally reduced by classified principal component analysis, so as to remove the correlation （information overlap） and mitigate the influence of multiple correlation on dimensionality reducing results. Finally, an integrated log is obtained from the filtered logs after dimensionality reducing, which improves the signal/noise ratio （SNR） and keeps thin-layer information efficiently through compounding the information of multiple logs. The method proposed in this paper is verified by the activity method and the inflection-point method which are simple and operable. Most thin layers can be automatically divided. The good application results reveal the efficiency and feasibility of this method in complex reservoir with thin interbeds.