Assessing the Reliability of Reduced-Retention-Time GPC for Hydrocarbon Fluid Analysis in Cruel Oil and Drilling Cuttings

This study explores the use of Gel Permeation Chromatography (GPC) combined with machine learning to characterize reservoir fluids from drilling cutting extracts. By reducing the GPC retention time by 50%, a more efficient workflow was tested using samples from viscous oil to gas condensate reservoirs in the NSC region. Statistical features derived from ultraviolet (UV) absorbance spectra of Storage Tank Oil (STO) samples were used to train a machine learning model for predicting API gravity. Approximately 20% of the STO dataset was reserved for validation, and the trained model was applied to cuttings-derived extract fluids. Results show that while the model can reasonably predict API for STO samples, uncertainties remain in predictions for drilling cuttings due to contaminants such as mud, coal, and wash fluids. The study highlights ongoing efforts to improve prediction accuracy by integrating multi-wavelength UV features and better calibration. Overall, this approach demonstrates the potential of GPC-based analytics for rapid and cost-effective reservoir fluid evaluation using drilling waste materials, supporting real-time geochemical interpretation and reservoir quality assessment.