There are a large number of methods and tools to analyze the quality indicators of natural gas. The most promising direction in the field is the correlation methods that use statistical models, in particular, neural networks. This is because it is possible to train these models. As a result, it is possible to get a clearer definition of the relationship between the model input and output parameters and to determine subsequently the required target parameters of natural gas. Currently lacks a general algorithm to determine the architecture and parameters of neural network models. Therefore, a comparative analysis of various models was previously carried out to solve the problem. Based on the results of such analysis, it was concluded that recurrent neural networks are main statistical models in this problem. But in contrast to the study mentioned above, this paper considers a recurrent neural network with a more complex architecture, namely, a network with a controlled recurrent block.