This paper proposes a method for metering natural gas energy based on the correlation between density and sound velocity. The technique integrates physical property correlation models with measured parameters, such as temperature, pressure, sound velocity, and density, to accurately predict the compression factor and the ideal volume calorific value of natural gas under operating conditions. The volume flow is corrected using the compression factor, which enables precise metering of natural gas energy through the adjusted volume flow and calorific value. To develop a high-precision physical property correlation model, a natural gas dataset comprising 10,000 sample sets is first constructed for model training and testing. Multiple machine learning algorithms are then employed to build predictive models. Analysis of the experimental results led to the development of a model-switching strategy based on the ranges of input features, which substantially enhanced prediction accuracy. For the compression factor model, the mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination (R2) were 0.00118, 0.0030, 0.14%, and 0.9987, respectively. The corresponding indicators for the calorific value model were 0.1583, 0.331, 0.44%, and 0.9736. The proposed method is finally validated using a natural gas real-flow test bench. The results demonstrated maximum prediction errors of 0.061% and 1.19% for the compression factor and calorific value, respectively, while the maximum relative energy error across four gas samples was 1.21%. These results indicate that the method can effectively achieve accurate natural gas energy metering in practical operating conditions.
Zhang et al. (Mon,) studied this question.