| Abstract |
Source rocks are fundamental elements for petroleum systems, and Total Organic Carbon (TOC) is one of the most important geochemical parameters in source rock property evaluation. The TOC determination methods using laboratory tests are expensive and limited, therefore prediction of TOC using geophysical well logs are vital for source rock characterization. Though there are various proposed TOC quantitation method, however, there still remains large uncertainty in delineation and quantitation of TOC using well log data due to the complex non-linear relationships between TOC and well log information, as well as the inherent limitations of various methods for TOC prediction. To fill the gaps between TOC and well logs, and eliminate uncertainties existing in empirical methods such as ΔlgR method, the geological, geophysical and geochemical data are integrated. History of source rock evaluation using well logs is reviewed, and sensitive well log parameters for source rocks are selected. The TOC content is correlated with well log series to unravel the well log responses of source rock intervals, and the organic matter rich intervals have high Uranium (U) concentrations and gamma ray (GR) readings, high sonic transit time (AC) and compensated neutron log (CNL), high resistivity, but low density readings. Then the various methods used for TOC quantitation are summarized in terms of their principles, interpretation process, and advantage and limitations. The Schmoker method is not applicable in shales, and borehole regularity will affect the linear regression relationship between TOC and bulk density. The Passey's ΔlgR method is widely used, however, the baseline selection will reduce the accuracy, and ΔlgR method is not applicable in highly mature or deep burial source rocks. The multiple regression analysis is hard to extend in other source rocks. The spectral GR method can hardly be used for lacustrine source rock analysis. The high acquisition costs of Nuclear Magnetic Resonance (NMR) and spectral mineral composition log (Schlumberger's Litho-Scanner logs) limit their extension in source rock evaluation. Artificial intelligence methods such as Back propagation (BP) neural network, Extreme Gradient Boosting (XGBOOST) can be used to predict TOC content via conventional logs, and the results are compared with the geochemical-measured TOC and ΔlgR method. The optimization of various methods for TOC prediction should fully consider their advantage and limitations. Additionally, comprehensive assessment of source rock should determine TOC, quality, and maturity of source rocks. This comprehensive review provides systematic and novel insights in applications of well logs in source rock evaluation, and has potential to fill gaps between geologists, geochemists and petrophysicists. |
| Authors |
Jin Lai  , Fei Zhao , Zongli Xia  , Yang Su , Chengcheng Zhang , Yinhong Tian , Guiwen Wang , Ziqiang Qin
|
