Abstract：

　　With the development of the oil industry to the deep water, underwater oil and gas production process have emerged and the traditional technology is facing many new problems. An alternative method for production estimation is represented by a Virtual Metering System (VMS) based on the analysis of the standard process parameters, available in almost all production system. The software is based on a methodology in which several models are included. This article mainly studies the application of an artificial neural network in gas well measurement. Because the existing wellbore models cannot adjust to changes of production in a timely manner nor predict accurately, this article introduced an error back propagation artificial neural network with highly nonlinear predictive ability. Artificially debugged wellbore model results served as a data sample library to simulate the mapping relationship between all kinds of influence factors and the gas well production.  A gas well flow calculation model based on a back propagation neural network is set up by learning and training. Predicted results show that compared with a physical flow meter, the average relative error of the calculation results is 3.33%. More than 80% of the data points have a relative error within plus or minus 5%, which indicates a high prediction accuracy. Comprehensive analysis shows that the artificial neural network model can meet the demands of practical production with the advantages of a simple model structure, flexible form and less calculation. Application of the artificial neural network model provides a new tool and method for virtual measurement technology.

Key words： subsea production system VMS Artificial Neural Network gas-condensate pipeline deepsea flow assurance

Received： 24 May 2017

Corresponding Authors：史博会, ydgj@cup.edu.cn; bh.shi@cup.edu.cn

Cite this article：SONG Shangfei,HONG Bingyuan,SHI Bohui等. Research into calculation of natural gas well production based on an artificial neural network[J]. Petroleum Science Bulletin, 2017, 2(3): 413-421.

URL：