Hydrate risk screening with CPA-based thermodynamics and surrogate models

Abstract

Natural gas hydrates, the crystalline, ice-like solids that form from the combination of light gas molecules and water under conditions of high pressure and low temperature, represent one of the most critical and persistent operational challenges facing the global energy industry. The formation and subsequent agglomeration of these solid masses into hydrate plugs within production tubing, subsea flowlines, transportation pipelines, and distribution networks pose a severe threat to flow assurance. These blockages can halt production entirely, leading to catastrophic economic losses measured in millions of dollars per day of downtime. Furthermore, the uncontrolled formation or dissociation of hydrates presents significant safety risks, making their prediction, prevention, and remediation a cornerstone of pipeline and process engineering. This thesis presents a comprehensive and integrated investigation into natural gas hydrate formation, employing a dual-pronged methodological approach. The primary objective is to bridge the gap between high-fidelity thermodynamic simulation and the practical, real-time needs of field operations. This study provides a rigorous assessment of existing thermodynamic models while simultaneously developing and verifying new, simplified analytical tools designed to empower engineers with the ability to make rapid and reliable risk assessments. The methodological framework of this research is founded on the synthesis of two distinct but complementary investigative pathways. The first component involved a deep, critical evaluation of the advanced thermodynamic packages available within the industry-standard OLGA dynamic multiphase flow simulator. This investigation specifically targeted the two primary thermodynamic engines used for hydrate prediction: the CPA Infochem model and the CPA Electrolyte model. The CPA Infochem model, which is based on the Cubic-Plus-Association (CPA) equation of state, is theoretically designed to excel in systems containing polar, associative components, as it explicitly accounts for the powerful hydrogen-bonding interactions characteristic of water, methanol (MeOH), and glycols. In contrast, the CPA Electrolyte model builds upon this foundation by incorporating terms to manage the complex, long-range electrostatic interactions introduced by dissolved inorganic salts. The performance of both models was meticulously benchmarked against a broad and diverse database of experimental hydrate equilibrium data collated from open academic literature. This testing matrix was extensive, covering the primary thermodynamic inhibitors used in the field: salt (Sodium Chloride, NaCl), MeOH, and monoethylene glycol (MEG), both individually and, crucially, in their binary-mixture combinations (e.g., NaCl +MeOH, NaCl +MEG). This benchmarking was intentionally performed for both single-inhibitor and mixed-inhibitor conditions to establish, at the outset of the study, the practical validity and applicability of the CPA-based hydrate frameworks across the inhibitor cases encountered in field operations. In this way, the subsequent analyses and correlation development were grounded on a simulation basis whose accuracy had been explicitly demonstrated with inhibition mechanisms. The second major pillar of this thesis was the development of novel analytical correlations. This initiative was driven by a core industrial need: the requirement for a predictive tool that is simple, computationally inexpensive, and rapid enough for daily use by field engineers and operators in settings such as urban gas distribution networks. As part of this effort, an analytical equation suitable for various salinity values was also derived, further simplifying the engineering calculations needed for rapid risk assessment. These operators often lack the time, specialized training, or software licenses required to run complex dynamic simulations. To build these correlations, a four-year, real-world operational dataset of natural gas compositions was secured from a major distribution company. This data provided a robust statistical basis for defining "average" and "boundary" gas profiles. Using the OLGA simulator, which had first been thoroughly verified against the experimental data, a large dataset of hydrate equilibrium points was generated for this average gas. This dataset was then subjected to rigorous statistical regression analysis, culminating in the derivation of new, explicit engineering equations (Equations 2.6, 2.7, and 2.8). These logarithmic equations were formulated to predict the precise hydrate formation temperature as a direct function of readily available field measurements: system pressure, water content, and salinity. The research findings provide clear and industrially significant insights into both inhibitor effectiveness and model performance limitations. The comparative assessment of inhibitor performance confirmed a distinct hierarchy of potency. As widely accepted, Sodium Chloride (NaCl) was re-confirmed as the most powerful thermodynamic inhibitor on a weight-percent basis. Its mechanism of action is fundamentally different from that of associative inhibitors; rather than merely competing with the hydrate lattice for hydrogen bonds, the powerful electrostatic fields of the Na+ and Cl- ions forcibly structure water molecules around themselves, sequestering them far more effectively than the hydrate cages can. When comparing the two primary organic inhibitors, the study observed that MeOH is, on a mass basis, marginally stronger than MEG. This slight thermodynamic advantage, however, is often negated by practical and economic realities. Methanol's high volatility causes significant losses to the gas phase, necessitating continuous replenishment and creating operational handling challenges. Conversely, MEG's extremely low volatility and its high capacity for recovery and reprocessing (i.e., regeneration) make it a far more sustainable, cost-effective, and operationally robust choice for long-term, closed-loop applications, such as those found in offshore production systems. A particularly important finding concerns the behavior of mixed-inhibitor systems (NaCl + MeOH and NaCl + MEG). Across the investigated concentration ranges, these mixtures produced a strong combined suppression of the hydrate equilibrium temperature, and the overall response was consistently close to additive. Within the scope of this study, no clear antagonistic (mutually weakening) behavior was observed. Minor departures from perfect additivity may be attributed to non-ideal interactions in the aqueous phase. This result has direct positive implications for the industry, suggesting that field operators can confidently deploy these mixtures, expecting a strong, combined, and largely predictable inhibitory response, even if the total effect deviates slightly from a perfect arithmetic sum due to the complex molecular competition for free water molecules. The critical assessment of the thermodynamic models themselves yielded crucial guidance for simulation engineers. The CPA Infochem model proved to be demonstrably superior in terms of robustness, stability, and predictive accuracy for the majority of systems tested. It performed exceptionally well in systems containing the polar, associative inhibitors (MeOH and MEG), and it maintained this high accuracy even in mixed systems containing low-to-moderate levels of salt. Its foundation in association theory appears to capably handle the dominant hydrogen-bonding physics. In stark contrast, the CPA Electrolyte model revealed significant limitations. While it performed adequately in systems containing only salt (high-salinity brines), its performance became inconsistent and unreliable in complex, mixed-inhibitor scenarios. It exhibited a clear and problematic tendency to "over-correct" the water activity calculation in the presence of both ions and glycols. This flaw caused the model to consistently over-predict the total inhibition effect, yielding results that are non-conservative and thus operationally unsafe. This strongly suggests that its parameters require significant recalibration to properly handle the simultaneous effects of ionic and associative forces. A final key observation applied to both models: at very high inhibitor concentrations (generally exceeding 30-40 wt%), both models began to show increasing deviation from the experimental data. This underscores a fundamental boundary of equilibrium modeling, where the system's behavior becomes dominated by other factors not accounted for in these equations. Arguably, the most significant and tangible contribution of this thesis is the successful development and verification of the new analytical correlations (Equations 2.6, 2.7, and 2.8). These equations, designed for simplicity and speed, were rigorously tested by comparing their predictions against the vast dataset generated by the verified OLGA CPA Electrolyte Model. The results of this verification were exceptional: the analytical correlations were able to reproduce the hydrate formation temperatures predicted by the complex simulator with an extremely low mean error margin, consistently falling within ±0.2 °C. This high degree of accuracy confirms their reliability for engineering purposes. The practical advantage conferred by this achievement is profound. These simple, validated equations completely eliminate the requirement for field personnel to access or operate complex, time-consuming, and expensive-to-license simulation software for routine risk assessments. They furnish engineers and operators with a powerful, reliable, and instantaneous tool to evaluate hydrate risk using only the basic operational data (pressure, water content, or salinity) that they already monitor in real-time. This work, therefore, successfully erects a practical and much-needed bridge between high-fidelity thermodynamic theory and the immediate, on-the-ground demands of field engineering. In conclusion, the findings of this thesis deliver direct, actionable, and economically valuable outcomes for two different operational domains with distinct needs: natural gas distribution networks and upstream gas production and flow assurance. For natural gas distribution lines, where rapid decision-making and real-time monitoring are essential, the newly developed analytical correlations provide a practical engineering solution that can be implemented in simple spreadsheet tools or integrated directly into SCADA (Supervisory Control and Data Acquisition) systems at city gate stations and distribution networks, enabling proactive, real-time hydrate risk warnings and preemptive operational adjustments. For gas production and subsea transportation systems, where fluids are multicomponent, and inhibitor selection/model fidelity drives both safety and cost, the detailed comparative assessment of inhibitors offers a scientifically grounded guide for selecting the most operationally effective and cost-efficient strategy (e.g., MEG for long-term subsea tiebacks versus MeOH for temporary interventions). In the same upstream context, the critical evaluation of the thermodynamic frameworks provides clear guidance for flow assurance design: the CPA Infochem model is recommended for systems containing alcohols or glycols to avoid the non-conservative behavior observed in the CPA Electrolyte model in complex mixtures. Beyond single-inhibitor scenarios, the thesis expands to multicomponent electrolyte systems containing both salt and thermodynamic inhibitors; alternative CPA-Electrolyte applications were compared, and a practical correction strategy was proposed to mitigate the excessive inhibition tendency in such mixtures. Finally, to support rapid screening without repeated full simulations (especially valuable for production-side sensitivity studies and operational envelopes), two surrogate models (Random Forest and Extreme Gradient Boosting) were trained on the Multiflash CPA Electrolyte model-generated.