Estimation of Uncertainties in Determination of Natural Gas Thermodynamic Properties Based Upon In-situ Raman Analyses of Gas Mixtures

Abstract

Raman spectroscopy is used in a real-time determination of the thermodynamic properties of natural gas. Two methods of Raman signal analysis are discussed. The first method is an analytical algorithm based on linear regression, which utilizes peak area. The second is an Artificial Neural Network algorithm that can utilize both peak area and its position. A detailed discussion of presented approaches is shown on reference data obtained using specially designed setup Raman with 105 m long fiber optic probe and a high-pressure cell. The proposed analysis allowed the calculation of composition for any arbitrarily chosen pressure. The feasibility of this approach is shown by the determination of composition for a gas flow in a separator installation on a gas production well. Results are compared with gas chromatography which allows the uncertainty determination of the method.