Low-Cost Empirical Methods for Predicting Water Content in Natural Gas System

Abstract

Accurate estimation of water content in natural gas is essential to prevent hydrate formation, internal corrosion, and to meet pipeline sales gas specifications. Direct measurement is often costly and not feasible for marginal fields, while rigorous thermodynamic models require licensed software and detailed composition. This study evaluates six low-cost empirical correlations for predicting water content in natural gas systems: McKetta-Wehe, Bukacek, Bahadori, Campbell, Towler-Mokhatab, and Katz models. The models, implemented in Microsoft Excel, were benchmarked against Aspen HYS V12.1 Peng-Robinson EOS predictions and published experimental data over pressure ranges of 200-3000 psia and temperature ranges of 60-160°F. Performance was assessed using Average Absolute Relative Deviation, Maximum Absolute Error, and R². Results show that the Bukacek correlation gave the best overall accuracy with AARD of 2.1%, followed by Towler-Mokhatab with 2.3% AARD, especially for sour gas containing CO₂ and H₂S. Bahadori’s model, though simplest, had the highest deviation of 5.6%. The study concludes that low-cost empirical correlations can provide sufficient accuracy for field design and operational decisions without dependence on expensive instrumentation or proprietary software. Recommendations are made on the appropriate selection of models based on pressure, temperature, and gas composition.