The development of natural gas quality parameters deter-
mination system have been studied. The proposed system is based on
neural network analysis and correlation analysis. The structure of the
developed system is presented. The main blocks and subblocks of the
proposed system are presented. The statistical model selection block and
the architecture and parameters of the model selection block that are the
most crucial stages of the proposed system are described in detail. The
mathematical principles of the discussed statistical models are provided.
The architectures of the studied neural network models are presented.
The results of comparative analysis of dierent statistical models includ-
ing neural networks are shown. The accuracy characteristics both for
training and testing stages of the neural networks are calculated. The
conclusion of nal neural network architecture for the studied task was
made. The results of testing of the proposed natural gas quality pa-
rameters determination system are provided. The steps for the further
research of the discussed task are considered.