Study on prediction model of liquid holdup based on back propagation neural network optimized by tuna swarm algorithm

ABSTRACT

It is unavoidable that there will be liquid accumulation in the low-lying areas of the pipelines during the operation of wet gas pipelines. The existence of liquid accumulation can generate a variety of safety issues and, in extreme circumstances, accidents. The accurate calculation of liquid holdup in gas-liquid two-phase flow is of great significance for the study of flow pattern identification, pressure drop calculation, pigging cycle determination, hydrate prediction, wax deposition prediction, pipeline corrosion evaluation and prediction, and transportation efficiency calculation of gas pipelines. Therefore, it is crucial to predict the liquid holdup of wet gas pipelines. 2141 independent experimental data samples were collected and screened out from literatures. Based on the gray theory, gray relation analysis was carried out on the influencing factors of liquid holdup, and the factors with greater influence were selected as the influencing variables; the liquid holdup prediction model based on tuna swarm algorithm optimized BP neural network was established, with pipe diameter, inclination angle, apparent gas velocity, apparent liquid velocity, average temperature, average pressure, and liquid viscosity as input parameters, and liquid holdup as output parameter. Liquid holdup was predicted for upward inclined, downward inclined, and horizontal pipelines respectively. The results show that the prediction model of liquid holdup established in this paper has high accuracy, with the MAPE value of 5.3223%, RMSE value of 0.0213, and R² value of 0.9924 for upward inclined pipelines; the MAPE value of 10.1859%, RMSE value of 0.0174, and R² value of 0.9922 for downward inclined pipelines; the MAPE value of 4.8037%, RMSE value of 0.0113, and R² value of 0.9974 for horizontal pipelines. The predicted results are generally stable and have a wider scope of application, providing a new idea and approach for predicting the liquid holdup of wet gas pipelines.

KEYWORDS:

• Wet gas pipeline
• Liquid holdup
• Grey relation analysis
• Tuna swarm optimization
• BP neural network

Disclosure statement

No potential conflict of interest was reported by the author(s).

Description of contributions of different authors

Rongge Xiao proposed and established the model in this paper, built the overall framework of the paper, and revised the paper and responded to reviewer comments; Guoqing Liu wrote the manuscript, created charts in the paper and performed the data analysis; Dongrui Yi designed computer programs and tested the data; Bo Liu and Qi Zhuang collected and screened data in literature, verified the accuracy and stability of the model.

Additional information

Notes on contributors

Rongge Xiao

Rongge Xiao is currently working as professor in the College of Petroleum Engineering, Xi’an Shiyou University, Xi’an, China. She received her PhD in Hydraulics and River Dynamics in 2014 from Xi’an University of Technology, Xi’an, China. Her research interests are multi-phase pipeline flow, optimization of natural gas treatment process, and flow assurance technology of oil and gas pipelines.

Guoqing Liu

Guoqing Liu is currently studying as a postgraduate in the College of Petroleum Engineering, Xi’an Shiyou University, Xi’an, China. He will receive his master's degree in Petroleum and Natural Gas Engineering in 2024 from Xi’an Shiyou University, China. His research interest is oil-gas storage and transportation engineering.

Dongrui Yi

Dongrui Yi is currently working as petroleum engineer in the Research Institute of Shaanxi Yanchang Petroleum (Group) Company Ltd., Xi’an, China. Her research interest is oil and gas gathering and transportation.

Bo Liu

Bo Liu is currently studying as a postgraduate in the College of Petroleum Engineering, Xi’an Shiyou University, Xi’an, China. He will receive his master's degree in Petroleum and Natural Gas Engineering in 2023 from Xi’an Shiyou University, China. His research interest is oil-gas storage and transportation engineering.

Qi Zhuang

Qi Zhuang is currently working as petroleum engineer in the Second Gas Production Plant, PetroChina Changqing Oilfield Company, Xi’an, China. He received his master's degree in Petroleum and Natural Gas Engineering in 2022 from Xi’an Shiyou University, Xi’an, China. His research interest is oil-gas storage and transportation engineering.