Accurate estimation of in-situ water saturation in unconventional shale reservoirs is critical for effective water production and management. Traditional resistivity-based water saturation calculation methods face challenges due to uncertainties in rock and fluid properties (e.g., water salinity, Archie parameters m and n), and usually need to be calibrated to core samples. This paper explores the relationship between triple-combo well logs and water-filled porosity from dielectric logs, considering factors like formations, well locations, and resistivity logging tool types. A new workflow that integrates physics and statistical methods is introduced to develop field-wide models for consistent water-filled porosity in wells with standard triple-combo logging suites.
Starting with Archie's equation, we establish the square root of electrical conductivity as the most effective transformation of resistivity for modeling water-filled porosity. We then perform exploratory analyses to identify additional factors affecting the relationship between water-filled porosity and resistivity, including formations and families of resistivity logging tools (induction or laterolog). Principle component analysis (PCA) is employed to visualize patterns in triple-combo logs and further explain the variance of water-filled porosity from dielectric logs. Upon experimenting with various combinations and configurations of PCA and multivariate regression, we select the best models and apply them to blind test wells. The study involves 20 wells from different basins.
We find that PCA and multivariate regression effectively model the relationship between water-filled porosity and triple-combo logs, with a margin of error under 1 p.u. across test wells. Interpreted formation tops improve prediction accuracy, while resistivity logs from induction and laterolog tools require separate treatment. Factors like well locations, mud types, and logging service providers show no significant impact. Overall, water-filled porosity based on statistical modeling achieves better consistency than conventional resistivity-based calculation across the dataset. Once learned with dielectric logs, models can be applied to wells with standard triple-combo logging suites for salinity-independent estimates.
This workflow combines the strengths of physics-based and data-driven models, offering a holistic view of the factors driving water saturation in unconventional formations. Due to the incorporation of domain knowledge and physical principles, the results proved to be robust and interpretable. The use of statistical methods helps discover underlying patterns in the log data and leverage information from a broader range of variables, leading to a more comprehensive understanding of the rock and fluid system and the relationships involved. The approach does not require prior knowledge of water salinity and is easily generalized to other fields.