A reliable model to predict mercury solubility in natural gas components: A robust machine learning framework and...

Mercury contamination in natural gas poses serious risks to production, processing, and transportation, leading to equipment corrosion, worker safety hazards, environmental pollution, and economic losses. Accurately predicting mercury solubility in methane, ethane, and multicomponent systems is essential for effective mitigation and regulatory compliance. This study employs advanced machine learning (ML) approaches, namely multilayer perceptron (MLP), generalized regression neural network (GRNN), and extra trees (ET), to estimate mercury solubility under varying pressure and temperature conditions. A high-quality dataset was used to train and validate these models, ensuring accuracy and reliability. The MLP model demonstrated the highest predictive performance with a determination coefficient of 0.9998, and a root mean square error of 1.7430 ppb. Besides, the MLP model effectively captured solubility trends, while feature importance analysis identified temperature as the dominant factor. The Leverage approach confirmed dataset reliability, with 96.5 % of data points within the trust region. This pioneering ML-based framework, the first of its kind for mercury solubility estimation, holds great industrial potential. It enables real-time monitoring, minimizes risks of equipment failure and human exposure, and supports environmental protection by reducing mercury emissions. By integrating this intelligent approach, operators can enhance safety, efficiency, and sustainability in natural gas operations.