Preicting CGS Pressure Using Machine Learning Case Study: Birjand

Sabzekar, Mostafa; Rezapour, Abbasali; Namakin, Motahare

doi:10.22052/eem.2025.255658.1091

Preicting CGS Pressure Using Machine Learning Case Study: Birjand

Document Type : Original Article

Authors

¹ Department of Computer Engineering, Birjand Universiy of Technology, Birjand, Iran

² Department of Civil Engineering, Birjand Universiy of Technology, Birjand, Iran

³ Department of Computer Engineering, Ferdowsi University of Mashhad, Mashhad, Iran

10.22052/eem.2025.255658.1091

Abstract

Today, natural gas stands as the dominant source of global energy supply. Iran holds the second-largest share of natural gas reserves, accounting for over seventeen percent of the world's total. However, it is concerning that gas consumption in Iran is three times the global average. To optimize planning in the natural gas supply sector, accurately assessing gas demand is crucial. Predicting natural gas consumption is vital for shaping energy policy and serves as a powerful tool for decision-makers, enabling them to effectively guide consumption and manage the balance between energy supply and demand. This paper addresses the challenge of predicting gas pressure at city gate stations (CGS) and analyzing its relationship with climate change. A significant issue faced in South Khorasan province during the colder months is the drop in pressure at these CGS, which stems from various factors, including the imbalance between gas production and consumption. Any estimates or forecasts regarding the pressure levels at these stations, which can serve as indicators of gas consumption, offer managers valuable insights to take proactive measures and mitigate potential crises. To tackle this problem, the study employs machine learning techniques. Data from CGS stations was sourced from the South Khorasan Province Gas Company, covering the years 2020 to 2024. Various scenarios were explored, including time series analysis, regression models, and the impact of temperature fluctuations on predictions, leading to the selection of the most effective approach. A notable strength of this research is the application of deep learning, a cutting-edge and highly promising machine learning methodology. Furthermore, this study marks the first instance of predicting CGS gas pressure. The findings underscore the significant influence of climate change factors on these predictions.

Keywords