The Utica-Point Pleasant formations in the Appalachian Basin cover an active area for the shale gas and oil exploration and development, especially in the eastern Ohio, western Pennsylvanian, and northwestern West Virginia. The total organic carbon (TOC) content, as one of the most important parameters describing the hydrocarbon potential of these shale formations, has been discussed a lot by various scholars in different shale plays. However, the behavior of TOC content in the Utica-Point Pleasant formations has been unanimously described to be different from the other shale plays, even though the Marcellus Shale is located in the same basin as the Utica-Point Pleasant formations. For example, not only the standard gamma ray log, but also the Uranium concentration from the spectral gamma ray log seems to lose its effectiveness in determining TOC content.
In this study, we will first investigate the best way to predict the TOC content in the Utica-Pleasant formation through combining core analysis data and wireline logs. Although the density log constitutes a good method for TOC prediction, the mineral composition of the shale matrix will affect the TOC content prediction. Thus, the mineral composition must be considered in this process. Also, the sensitivity of various logs will be evaluated for the TOC prediction, including the gamma ray, neutron, acoustic log, and resistivity log, which reflect the mineral composition information. Different mathematical methods, such as multiple regression, artificial neural network, support vector machine, and differential evaluation, will be employed and compared in determining the best models for the TOC prediction.
A large dataset of wireline logs for the Utica-Point Pleasant formations is already available. With a reliable prediction of TOC in the wells with wireline logs, a 3-D model of TOC content could be constructed. The geostatistical algorithms, such as kriging as a deterministic method and the Gaussian Sequence Index as a stochastic method, have significant influence on the TOC distribution modeling, and several TOC models using various geostatistical algorithms are compared to figure out the best one. The spatial distribution of the TOC content within the organic-rich shale thickness will show very useful information for detecting the sweet spots in shale plays.