Electrical Conductivity Characterization in Shaly Reservoir Rocks Through Measurements and Optimization Techniques

ABSTRACT

Cation‐exchange capacity per volume (QV) plays a pivotal role in reservoir engineering, influencing ion transport, wettability alteration, and enhanced oil recovery (EOR) processes. This study introduces a gradient boosting machine (GBM) framework integrated with four optimization strategies, including evolution strategies (ES), batch Bayesian optimization (BBO), Bayesian probability improvement (BPI), and Gaussian process optimization (GPO), to achieve robust characterization of QV in oil and gas applications. Comparative evaluation based on test‐stage performance demonstrated consistently high accuracy, with GBM‐ES and GBM‐BPI achieving superior R ² values (0.987) and low mean squared errors (0.00062 and 0.00063, respectively). GBM‐BPI further minimized average absolute relative error (2.776%), whereas GBM‐ES balanced accuracy with computational efficiency. Runtime analysis revealed trade‐offs among optimizers, with GPO converging fastest (532.5 s) compared to BBO's extended runtime (1537.9 s). SHapley Additive exPlanations (SHAP)‐based interpretability and sensitivity analysis provided transparent insights into feature contributions, identifying the dominant reservoir parameters governing QV variation. By emphasizing test‐stage performance and interpretability, this framework establishes a reproducible methodology for QV assessment, offering a scalable tool for reservoir characterization, ion‐exchange modeling, and optimization of EOR strategies in the petroleum industry.