NEURAL NETWORK MODELING OF ATMOSPHERIC AIR POLLUTION BY MAIN GAS PIPELINES COMPRESSOR STATIONS

Abstract

Background An important role in solving the problems of technical diagnostics is assigned to prompt, reliable monitoring of the operational parameters of the process equipment. At the same time, the existing information-measuring systems should be supplemented by computational methods of technological parameters of operation of gas pumping units (GPU). Aims and Objectives It is known that the amount of emissions of nitrogen oxides and carbon with exhaust gases of GPU correlates with the processes of aging and deterioration of aggregates. The task of developing computational methods for modeling the amount of emissions by GPU as an argument of the GPU technical state is relevant. Methods The database for solving the task can be provided with a permanent system for measuring the technological parameters of operating units. In world practice, automated monitoring systems are used to control emissions of harmful substances. However, the high cost of such equipment in the oil and gas market prevents their widespread use. It is shown that the concentrations of nitrogen and carbon oxides in the GPU exhaust gases have a significant correlation with the temperature of the gases, the rotation frequency and the air temperature at the entrance to the axial compressor. This fact proves the need for multi-parameter modeling to determine the concentration of nitrogen oxides and carbon. The solution of practical important problems of complex technological processes is recommended within the neural network model implementation. Results Based on the values of the main technological parameters of the GPU, neural network models for estimating emissions of nitrogen oxides and carbon from the GPU-10 Volna aggregates with an error of less than 7 % were constructed.