Analysis of natural gas: the necessity of multiple standards for calibration

The importance of natural gas as an international trading commodity and the cost to consumers has made the accuracy of determinations for the components of natural gas very important. Pricing of natural gas is based on the heating value of the gas determined from either calorimetry measurements or calculations based on individual component concentrations determined by gas chromatography (GC). Due to the expense of accurate calibration standards, many analysts and laboratories will use a single calibration standard to perform natural gas determinations. Therefore, the purpose of this study was to determine whether an analyst could accurately measure the components of natural gas, in particular methane, using a single standard, or whether a suite of standards is necessary to calibrate the analytical instrument. A suite of eight gravimetric primary standards was prepared covering a concentration range for methane of 64-94 mol%, with uncertainties of +/-0.05% relative (95% confidence interval). The natural gas primary standards also contained nitrogen, carbon dioxide, ethane, propane, iso-butane, n-butane, iso-pentane, n-pentane, and n-hexane with varying concentrations from 0.02 to 14%. A single analytical method was used in which only the amount of sample injected onto the column was altered. The results show that when injecting a 0.5 ml sample volume a second-order regression through the standards is necessary for the determination of methane. The results for nitrogen, ethane and propane also, show the same trend. Only those individual standards whose methane concentration is within 1% of the test mixture predicted a concentration within 0.05% of the regression value. Those individual primary standards whose methane concentration is different by more than +/-1% of the test mixture predicted values differing by +/-0.5 to +/-2.0% from the regression value. These differences lie well outside the predicted concentration uncertainty interval of +/-0.20%. A smaller sample volume 0.1 ml resulted in a set of data that could be fit using linear regression. Each of the eight primary standards individually predicted the methane in the test mixture to be within +/-0.11% of the predicted value from linear regression. The data confirm that it is imperative to fully characterize the analytical system before proceeding with an analysis. (C) 2003 Elsevier B.V. All rights reserved.