Data-driven modelling for gas consumption prediction at City Gate Stations

Abstract

Abstract City Gate Stations (CGS) are critical elements of Natural Gas (NG) distribution systems, as they connect national high-pressure transmission networks with local low-pressure networks. One of their main tasks is to pre-heat the gas to avoid dangerous sub-cooling due to the Joule-Thompson effect after the pressure reduction stage. For this process, significant amounts of thermal energy are required, usually obtained by burning part of the gas flow rate. This work aims to develop a data-driven model that will serve as a tool to predict and monitor the thermal consumption of the CGS. The plant chosen as a case study for this activity is in a region of central Italy. A Multiple Linear Regression (MLR) model is developed and trained, and its predictive performance is evaluated. The model results achieved an accuracy of over 95% for the coefficient of determination. The method makes it possible to create a baseline consumption model and evaluate the performance of the CGS over time using the CUSUM technique to find variations between actual and modelled gas consumption, being an essential tool for monitoring the effectiveness of the NG preheating system.