Determining the precise amount of natural gas (NG) volumetric flow rate is an important issue in trading NG. The temperature, pressure, and compressibility factor of NG are required parameters for calculating (NG) volumetric flow rate. To calculate the NG compressibility factor according to the acceptable standard methods, the NG composition analysis must be known. Considering that measuring NG composition analysis is a costly and complex method, an artificial intelligence (AI) method has been presented to calculate the NG compressibility factor by measuring simple thermodynamic properties and without any information about NG composition analysis. Temperature, pressure, density, and heating value have been selected as measurable properties as well as input parameters to the neural network models. A large database containing 50,000 distinct natural gases with thermodynamic properties has been generated using the GERG-2008 equation of state (EoS) to train the proposed neural network model. The proposed neural networks have been validated by using available experimental data. The statistical parameters show that the 4-input network calculates the compression factor with higher accuracy than the 3-input network. Average absolute percent relative error, root mean square error, and coefficient of determination have been reported 0.2%, 0.0028, and 0.9943 for 4-input network compared to experimental data. Moreover, the two proposed neural networks have been presented for NG samples collected from different regions of Iran and the results compared to the GERG-2008 EoS. The results show that the proposed neural network method calculated the NG compressibility factor with acceptable precision.

Keywords