摘要
Energy efficiency and energy savings are priority activities to OAO Gazprom and represent a set of policy measures aimed at the rational use and saving of energy resources, an important part of which is forecasting and planning of electricity consumption at compressor stations (CSs). The main consumers of electricity at compressor stations with gas-turbine gas-compressor units are electric motors with a short-circuited rotor of air cooled heat exchanger (ACHE) of gas. This article analyses the raw data at the Petrovsk–Pisarevka site of the Urengoi–Novopskov pipeline (the installed capacity of ACHE at each CS, the carrying capacity of the CS depending on the operation mode of the ACHE, the consumption of electricity during the year per month, and the temperature at the inlet and outlet of the CS). The relationship between such variables as the electricity consumption, the carrying capacity of the main gas pipeline, and the gas temperature at the outlet of the CS is found. To solve the problems that have been posed, we used the method of neural networks, which has a number of advantages over the regression models: such networks on their own select the form of the functional dependence in accordance with experimental data and are an adaptive model that adjusts the structure of the network for new observations and can explain rather complex relationships between the energy consumption and indicators of the main gas pipeline. Keywords main gas pipeline compressor station air-cooled heat exchanger (ACHE) of gas prediction mathematical statistics neural networks