Robust pressure prediction of oil and gas pipeline networks based on equipment embedding neural network

Abstract

Currently, extensive pipeline networks are developed in response to the demands of the oil and gas industry. The accurate estimation of the hydraulic condition of pipeline networks holds significant importance in the fields of pipeline design and safety management. Nevertheless, predicting the pressure of oil and gas pipeline networks with different equipment and structures remains challenging. To meet this challenge, a novel pressure prediction model for the oil and gas pipeline networks based on the equipment embedding neural network (EENN) is proposed in this study. The proposed model embeds different equipment models into the neural network model. The neural network in this model is used to focus on learning the connection characteristics of the pipeline network to achieve higher prediction accuracy. The present study first explores different embedding combinations of the EENN model to estimate the pressure in an oil pipeline network system that involves a non-isothermal batch transportation process. Then, the trained model is applied to predict the pressure in a gas pipeline network. The optimal EENN exhibits an average prediction error of 18.5% for oil pipelines and 0.36% for gas pipelines, which is lower than 20.8% and 3.57% under the neural network. The findings of this study demonstrate the efficacy of the proposed EENN in accurately forecasting pressures in diverse oil and gas pipeline networks by reducing the complexity of the learning process.