Fourier Transform Near-Infrared (FTNIR) Spectroscopy and Partial Least-Squares (PLS) Algorithm for Monitoring Compositional Changes in Hydrocarbon Gases under In Situ Pressure

The ability of Fourier transform near-infrared (FTNIR) spectroscopy and chemometric method was investigated to determine the concentration of major hydrocarbon components of natural gases at pressures from 3.44 to 13.78 MPa and temperatures from 278.15 to 313.15 K. Various partial least-squares (PLS) models were developed to determine the concentration of methane, ethane, propane, i-butane, and n-butane simultaneously in gas phase at different pressures and temperatures, using the acquired FTNIR spectra. Several preprocessing techniques were tested prior to the construction of the calibration models. The first Savitzky−Golay derivative with smoothing over five points plus orthogonal signal correction (OSC) was found to be the best method for FTNIR data preprocessing. Good agreement was obtained between the predicted data by PLS models and the measured values with a standard error of prediction (SEP) of 0.184−0.217, 0.165−0.209, 0.136−0.181, 0.098−0.154, and 0.096−0.142 cmol/mol for methane, ethane, propane, i-butane, and n-butane, respectively, under different temperature and pressure conditions. The developed PLS models were evaluated for a real natural gas, and a good agreement between the PLS model prediction and the gas chromatography (GC) analysis was gained at different pressures. Finally, the sensitivity of the FTNIR spectroscopy technique to the system pressure and temperature was investigated. It was verified that changes in pressure and temperature within a certain range affect the accuracy of the PLS models. The results suggest that FTNIR spectroscopy in association with chemometric method, based on the PLS algorithm, is a viable approach for monitoring changes in the concentration of major components in the gas phase.