Refined Lithological Classification through Structured Multivariate Analysis

Estimation of reservoir heterogeneity from well logs is an important yet difficult task encountered in geophysical formation evaluation and reservoir engineering. This paper presents a methodology for mapping petro-variability through a structured multivariate analysis of four wells from an Upper Triassic to Lower Jurassic fluvial succession, offshore mid-Norway. Traditionally principal component analysis (PCA) is run by analyzing the entire wireline log and using PCA scores to characterize variability within and between lithologies. In this paper we propose a technique to quantify reservoir heterogeneity due to second order rock property contrasts in addition to more standard approaches for lithological segmentation. In this study, facies characterization has been addressed through the use of four separate PCAs. PC loadings and scores are calculated from: (1) a total analysis of all facies and all log variables in addition to derived variables through a 1st and 2nd derivative transform, (2) a subset analysis including only the coaly facies and (3) a second subset analysis including only cemented zones and finally (4) a refined reservoir heterogeneity analysis portraying the effects of secondary sources of petro-variability is quantified through the use of the Eckart-Young theorem. This complex reconstruction of the log variables into petro-normalized logs can then be used to portray the within-facies heterogeneities and refine the more standard characterization of reservoir heterogeneity. This paper demonstrates that the petro-normalization using the Eckart-Young theorem enhances our ability to characterize reservoir heterogeneity compared to the more traditional workflows. The methodology has been successfully tested on a cored well and the application of this procedure provided additional insight into the complex fluvial characteristics of a heterolithic sequence that poses great development challenges.