Mechanical Prediction of Shale Oil Reservoirs Based on Micromechanical Characterization, Machine Learning, and...

Shale oil reservoirs exhibit strong heterogeneity and complex rock mechanical properties. Accurate geomechanical modeling is crucial for optimizing horizontal well trajectories and designing fracturing parameters. Traditional mechanical experiments are costly and provide limited data, while empirical formulas suffer from poor regional applicability. Machine learning models designed to predict reservoir mechanical parameters commonly overlook lithofacies-specific variations. To overcome this limitation, this paper presents and adopts a novel workflow that combines micromechanical characterization, machine learning prediction, and mechanical upscaling. Nanoindentation tests are first performed in Well W1 in the Ordos Basin to obtain the mechanical properties of different micro-constituents in shale. Then, a machine learning model is used to establish the relationship between conventional well logs and micro-constituent contents, thereby predicting the distribution of micro-constituents along the entire wellbore. Finally, based on micromechanical derivation, the Mori-Tanaka model is applied for mechanical upscaling to obtain the Young's modulus. The feasibility of this method is validated by comparing the results with those calculated from conventional well log data. This approach combines the high-precision micromechanical characterization of nanoindentation with the predictive capability of machine learning. It overcomes the limitations of elemental capture spectroscopy logging (namely, its high cost and insufficient organic matter information) and provides a new pathway for establishing one-dimensional geomechanical models.