Re-Calibration of Empirical PVT Correlations for Crude Oil Formation Volume Factor Using Non-Linear Optimization Algorithms

Abstract

The oil formation volume factor (Bo) is one of the most critical parameters in petroleum reservoir engineering, governing reserve estimation, material balance calculations, well deliverability assessment, and production facility design. Widely adopted empirical correlations, principally those of Standing [26] and Glaso [13], were derived from geographically limited datasets and may exhibit systematic bias when applied to crude oils from geologically distinct basins, including West African and offshore deepwater reservoirs. This paper presents a numerical optimization study aimed not at deriving new correlations, but at recalibrating the existing coefficients of Standing's and Glaso's Bo correlations to better represent a broader PVT dataset comprising 567 laboratory-measured samples. Multiple optimization algorithms including Nelder-Mead simplex, Broyden-Fletcher-Goldfarb-Shanno (BFGS), Limited-memory BFGS with bounds (L-BFGS-B), Conjugate Gradient (CG), Powell, and Differential Evolution were employed to minimize the mean squared error between measured and predicted Bo values. Results demonstrate that coefficient recalibration achieves statistically significant improvements across performance metrics. For Standing's correlation, the optimized form reduces RMSE from 0.1503 to 0.1336 bbl/STB, representing an 11.1% reduction. For Glaso's correlation, RMSE reduces from 0.1528 to 0.1350 bbl/STB, representing an 11.7% reduction. Five-fold cross-validation confirms the generalization capability of the recalibrated coefficients. This work provides a transparent and reproducible framework for PVT correlation recalibration applicable to regional datasets, offering practicing engineers a low-cost alternative to full correlation redevelopment.