Intelligent solubility estimation of gaseous hydrocarbons in ionic liquids

Abstract

The research focuses on evaluating how well new solvents attract light hydrocarbons, such as propane, methane, and ethane, in natural gas sweetening units. It is important to accurately determine the solubility of hydrocarbons in these solvents to effectively manage the sweetening process. To address this challenge, the study proposes using advanced empirical models based on artificial intelligence techniques like Multi-Layer Artificial Neural Network (ML-ANN), Support Vector Machines (SVM), and Least Square Support Vector Machine (LSSVM). The parameters for the SVM and LSSVM models are estimated using optimization methods like Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Shuffled Complex Evolution (SCE). Data on the solubility of propane, methane, and ethane in various ionic liquids are collected from reliable literature sources to create a comprehensive database. The proposed artificial intelligence models show great accuracy in predicting hydrocarbon solubility in ionic liquids. Among these, the hybrid SVM models perform exceptionally well, with the PSO-SVM hybrid model being particularly efficient computationally. To ensure a comprehensive analysis, different examples of hydrocarbons and their order are included. Additionally, a comparative analysis is conducted to compare the AI models with the thermodynamic COSMO-RS model for solubility analysis. The results demonstrate the superiority of the AI models, as they outperform traditional thermodynamic models across a wide range of data. In conclusion, this study introduces advanced artificial intelligence algorithms such as ML-ANN, SVM, and LSSVM in accurately estimating the solubility of hydrocarbons in ionic liquids. The incorporation of optimization techniques and variations in hydrocarbon examples improves the accuracy, precision, and reliability of these intelligent models. These findings highlight the significant potential of AI-based approaches in solubility analysis and emphasize their superiority over traditional thermodynamic models.