A Big Data Method Based on Random BP Neural Network and Its Application for Analyzing Influencing Factors on Productivity of Shale Gas Wells

• Published in: Energies (Basel) Vol. 15; no. 7; p. 2526
• Main Authors: Zhao, Qun, Zhang, Leifu, Liu, Zhongguo, Wang, Hongyan, Yao, Jie, Zhang, Xiaowei, Yu, Rongze, Zhou, Tianqi, Kang, Lixia
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.04.2022
• Subjects: Artificial intelligence; Back propagation networks; Big Data; big data analysis; BP neural network; Deep learning; Engineering; gas well productivity; Geology; Geophysics; Machine learning; Network analysis; Neural networks; Oil exploration; Productivity; random probability; Shale; Shale gas; Statistical analysis; Statistical methods; Technology; China
• ISSN: 1996-1073; 1996-1073
• DOI: 10.3390/en15072526

In recent years, big data and artificial intelligence technology have developed rapidly and are now widely used in fields of geophysics, well logging, and well test analysis in the exploration and development of oil and gas. The development of shale gas requires a large number of production wells, so big data and artificial intelligence technology have inherent advantages for evaluating the productivity of gas wells and analyzing the influencing factors for a whole development block. To this end, this paper combines the BP neural network algorithm with random probability analysis to establish a big data method for analyzing the influencing factors on the productivity of shale gas wells, using artificial intelligence and in-depth extraction of relevant information to reduce the unstable results from single-factor statistical analysis and the BP neural network. We have modeled and analyzed our model with a large amount of data. Under standard well conditions, the influences of geological and engineering factors on the productivity of a gas well can be converted to the same scale for comparison. This can more intuitively and quantitatively reflect the influences of different factors on gas well productivity. Taking 100 production wells in the Changning shale gas block as a case, random BP neural network analysis shows that maximum EUR can be obtained when a horizontal shale gas well has a fracture coefficient of 1.6, Type I reservoir of 18 m thick, optimal horizontal section of 1600 m long, and 20 fractured sections.