Forecasting the Occurrence of Drop Pressure and Over Pressure of Natural Gas Pipeline Distribution Network Using Machine Learning

Maintaining stability in natural gas distribution is essential for ensuring a reliable energy supply and supporting sustainable energy goals. Real-time monitoring and simulations are crucial for managing supply disruptions that can cause dangerous pressure fluctuations, which, if not addressed promptly, can lead to costly production delays and significant safety risks. Overpressure conditions can push systems beyond safe operational limits, threatening infrastructure, and personnel. Stabilizing the distribution network not only addresses safety concerns but also improves energy efficiency, reduces waste, and supports sustainability efforts. This study evaluates four predictive models: ARIMAX, SARIMAX, random forest regression, and linear regression to mitigate risks and enhance operational decision making in natural gas distribution. The models are assessed using key performance metrics, including mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE), to identify the most accurate and reliable method for real-time predictions. Given the operational and financial risks associated with supply imbalances, selecting the right predictive model is critical for maintaining network stability and minimizing potential losses. The findings indicate that random forest regression provides the highest accuracy in real-world testing, making it the most suitable model for predicting complex pressure fluctuations. In contrast, linear regression was less effective due to its higher variability when handling the complexities of pressure management. Ultimately, choosing the appropriate predictive model is key to ensuring network stability, reducing risks, and promoting sustainable energy practices.