Enhancing Prediction Accuracy of Gas Viscosity Correlations for the Niger Delta Basin by Parameter Calibration using Nonlinear Optimization Algorithms

Abstract

Accurate prediction of natural gas viscosity is essential for reliable reservoir engineering calculations, fluid flow analysis, and production system design. Although laboratory measurements provide accurate viscosity values, they are often expensive and time-consuming, leading to widespread reliance on empirical correlations. However, commonly used industry-standard correlations developed from global datasets often fail to accurately represent the thermophysical behavior of natural gas systems in the Niger Delta Basin, while existing region-specific correlations show limited accuracy, and their accuracy has not been optimized through model calibration techniques. In this study, a nonlinear least squares model fitting technique built on the Levenberg–Marquardt optimization algorithm via Python’s scipy.optimize.curve_fit was used to optimize the Lee-Gonzalez-Eakin correlation and the Carr-Kobayashi-Burrows correlation (modified by Dempsey and Standing), using 5,511 experimental PVT data points obtained from Niger Delta gas fields. The dataset was divided into training, testing, and validation datasets to assess performance robustness. Model performance was evaluated using the coefficient of determination (R²), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). The calibrated models showed good improvements in predictive accuracy. The R² increased significantly for both Lee-Gonzalez-Eakin and Carr-Dempsey-Standing correlations from approximately 0.88 to greater than 0.9999. The MAE and RMSE for the Lee-Gonzalez-Eakin correlation reduce to zero from 0.001310 and 0.001315, while the Carr-Dempsey-Standing correlation also reduces to zero from 0.001014 and 0.001327. In contrast, the existing region-specific correlation developed by Azubuike et al. for the Niger Delta Basin gave a moderate R² of 0.58025 and was significantly outperformed by the calibrated global models. The results show that while global correlations for gas viscosity provide a useful starting point, model calibration using region-based datasets significantly enhances their accuracy and reliability for petroleum engineering applications in the Niger Delta.