Zhijun Pei, PhD Candidate, China University of Petroleum, Beijing, presents his talk "Intelligent Optimization of Drilling Efficiency Based on Prior Physical Knowledge".
Abstract: Drilling engineering uses mechanical equipment to break formation rocks and form communications with subsurface resources. However, the high cost of drilling directly affects the large-scale development and utilization of underground resources, such as petroleum, geothermal, and CO2 storage in deep formations. Increasing the drilling efficiency can effectively reduce drilling costs, solving the outstanding issues in the uneconomical cost of geothermal and CO2 storage in deep formations. In recent years, the development of artificial intelligence technology has set off the fourth industrial revolution, and intelligent drilling technology has also emerged, which is also considered a transformative technology that can significantly reduce drilling costs. However, the accuracy, stability, and mobility of fully data-driven artificial intelligence models are difficult to meet the requirements of field applications. Therefore, exploring a reliable and accurate drilling optimization method that combines prior physical knowledge with artificial intelligence technology has become the key to intelligent drilling technology.
In this study, an intelligent rate of penetration (ROP) prediction model embedded with prior mechanism knowledge was established from physical boundary conditions, neural network structure, and control equations, and the established proxy models were analyzed and explained through interpretable methods. Based on multimodal graph learning, multiple individual drilling agents are fused to achieve a more accurate representation of the drilling environment under limited data conditions. Use intelligent optimization algorithms to achieve multi-objective collaborative optimization of drilling parameters and automatically generate and optimize the drilling scheme.
Speaker Bio: Zhijun Pei is a PhD candidate at China University of Petroleum, Beijing, under the supervisor of Dr. Xianzhi Song and will defend his PhD in June 2024. He has published 5 journal papers and authorized 3 patents for his PhD work. He also is the Youth editorial board member of Xinjiang Petroleum and Natural Gas Journal. He spent his last year at the University of Alberta under the supervision of Dr. Bo Zhang and continued his research on using artificial intelligence (AI) techniques to achieve reliable prediction and optimization of the drilling efficiency and cost for subsurface energy development.
12 окт 2023
