Abstract: “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.

Key words: shale oil, thin layer, principal component analysis, multiple correlation, automatic layering, activity method, inflection-point method