Mechanism-Constrained Electrical and Non-Electrical Log Fusion for Oil Saturation Evaluation in Low-Resistivity Lacustrine Shale Oil Reservoirs

Accurate oil saturation evaluation is critical for shale oil sweet-spot prediction and reserve assessment, but conventional resistivity-based interpretation is commonly challenged in clay-rich lacustrine shale oil reservoirs. In some favorable shale intervals, high oil-bearing potential is accompanied by low resistivity, which weakens the applicability of Archie-type saturation models and may lead to underestimation of oil saturation. To clarify the origin of this abnormal electrical response and improve saturation interpretation, this study investigates the upper fourth member of the Shahejie Formation in the Boxing Sag, Jiyang Sub-basin. Core description, thin-section observation, XRD mineral composition, porosity, measured oil saturation, TOC, S1 and conventional logging data were integrated to establish a core-calibrated and SOM-based reservoir classification workflow. Using GR, AC, CNL, DEN, Rt and Rs as input curves, the shale oil reservoirs were divided into Type I, Type II and Type III classes. Type I reservoirs exhibit better pore development and hydrocarbon enrichment, with mean porosity, oil saturation, TOC and S1 values of 4.95%, 59.18%, 2.21 wt% and 2.77 mg/g, respectively. However, their mean Rt is only 22.75 Ω·m, much lower than that of Type III reservoirs. This response defines a typical “three-high and one-low” logging pattern, characterized by high GR, high AC, high CNL and low resistivity. The resistivity-reversal behavior is mainly associated with clay-related conduction and relatively higher absolute water content, whereas organic matter and pyrite do not form continuous conductive pathways in the studied interval. Based on this mechanism, a class-dependent electrical–non-electrical log fusion model was developed by combining clay-corrected electrical saturation with neutron–density saturation constraints. Compared with the single electrical model and single non-electrical model, which yield mean absolute errors of 12.65 and 13.84 percentage points, respectively, the proposed fusion model reduces the error to 8.88 percentage points. Application to more than 40 wells identifies the central–western part of the Boxing Sag as the most favorable shale oil sweet-spot area. The cumulative thickness of Type I reservoirs generally exceeds 100 m in this area and locally reaches more than 150 m. The proposed workflow provides a practical method for identifying favorable low-resistivity lacustrine shale oil reservoirs using conventional logging data.