Objective The linear stimulation and covariance calculation of multiple sets of data are used to study the correlation of natural gas composition analyzed by gas chromatography and its influence on the uncertainty of calculated physical properties.

Method More than 40 times of composition analysis on each of 5 natural gas samples were put together to get the correlation coefficients between different compositions in samples. Five correlation coefficient matrixes were formed according to these correlation coefficients and their average value was calculated. Combined with the characteristics of analysis equipments and analysis process, the causes of correlation coefficient matrix were discussed, and the method was recommended to determine the correlation coefficient in the situation that the correlation coefficient cannot be practically measured.

Result Based on the simulated natural gas compositions and their uncertainties, a relative 20% of difference between the uncertainties of natural gas calorific value was observed using different matrix. The influence on uncertainty evaluation was analyzed by the difference of correlation coefficient.

Conclusion The more overlapping between difference compositions in the process of sampling, separation, and detection, the stronger correlation was found. The larger correlation coefficients, the larger uncertainty of the calculated physical parameter.