On determination of natural gas density: Least square support vector machine modeling approach

摘要: Abstract In this century, worldwide consumption of natural gas is expected to increase drastically because it one the cleanest and most available energy sources. Accurate knowledge properties a vital significance in engineering. One important density, which traditionally measured through expensive, time consuming cumbersome experiments. communication, new reliable accurate model for prediction density presented as function pseudo reduced pressure, temperature apparent molecular weight gas. A supervised learning algorithm, namely least square support vector machine, has been employed modeling parameters were optimized coupled simulated annealing. The results study indicated that developed can satisfactorily predict wide range pressure (from 13.7 10,000 psia), (from −25 460 °F) composition (molecular from 16.04 129.66). Moreover, accuracy validity proposed was compared pre-existing models, found more accurate, superior all investigated models. addition, relevancy factor demonstrated greatest impact on among selected input parameters.