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

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