Forecasting Gas Pipe Pressure Using ANN and Linear Regression

In the oil and gas industry, pipelines are essential for gas transport. Pipelines make it easier for companies to transport gas over long distances. However, sometimes pipeline failures occur and are generally unexpected. Therefore, a method is needed to carefully monitor gas transportation pipelines to minimize the occurrence of pipeline failures. One of the things that affects gas transportation is the pressure in the pipeline so that the gas in the pipeline reaches its destination. This research aims to apply machine learning methods to forecast gas pipeline pressure using Artificial Neural Network (ANN) and Linear Regression methods, utilizing data collected in 2021 that records gas pipeline conditions per hour. By using historical pressure data to predict future outcomes, we explore the ability of ANN and Linear Regression models in this forecasting task. Linear Regression is chosen for its compatibility with numerical data, while ANN is chosen for its ability to capture complex patterns within numerical datasets. The performance of these methods is evaluated through error value metrics, ensuring accurate pressure forecasts. The Linear Regression that forecasted six hours has error metrics of 0.39477, 0.62831, and 0.46523 psig using MSE, RMSE, and MAE. While a four-layer ANN resulted in an error metric of 0.34596, 0.58819, and 0.43158 psig using MSE, RMSE, and MAE. The findings indicate that the four-layer ANN model outperforms the Linear Regression model in pressure forecasting accuracy. This research contributes to enhancing the reliability of gas pipeline pressure forecasting, thereby aiding in the prevention of pipeline failures and ensuring efficient gas transportation operations.