压裂泵注智能决策系统设计理念与路径探讨  

Design concept of intelligent decision-making system for hydraulic fracturaing pumping and its implementation path

作　　者：盛茂[1,2] 田守嶒[1,2] 朱丹丹 王天宇 廖勤拙 李根生[1,2] SHENG Mao;TIAN Shouceng;ZHU Dandan;WANG Tianyu;LIAO Qinzhuo;LI Gensheng(State Key Laboratory of Oil&Gas Resource and Engineering//China University of Petroleum-Beijing,Beijing 102249,China;College of Artificial Intelligence,China University of Petroleum-Beijing,Beijing 102249,China)

机构地区：[1]油气资源与工程全国重点实验室·中国石油大学(北京) [2]中国石油大学(北京)人工智能学院

出　　处：《天然气工业》2024年第9期108-113,共6页Natural Gas Industry

基　　金：国家自然科学基金优秀青年科学基金项目“油气井流体力学与工程”(编号:52122401);国家自然科学基金重点国际(地区)合作项目“水力压裂多尺度多场耦合问题的智能表征理论与方法”(编号:52320105002)。

摘　　要：油气藏改造过程中的压裂泵注具有工艺复杂多变、过程不可逆、高风险等特点,需要及时准确地进行决策调控,进而提升压裂改造效果。目前决策方式主要以压裂设计泵注程序为基础,人为调控泵注参数。随着人工智能技术的发展,构建压裂泵注智能决策系统、实现自主决策已成为可能。为此,在综述国内外压裂泵注决策研究现状的基础上,提出了基于强化学习理论的压裂泵注智能决策设计理念,并探讨了实现路径。研究结果表明:(1)压裂泵注智能决策系统的核心是泵注决策智能体,由其自主决策泵注程序和优化参数,同时将决策动作传递给地面装备,构成闭环调控,实现泵注过程“自动驾驶”;(2)基于强化学习理论的压裂泵注智能决策设计理念,即以强化学习为核心,以压裂大数据为基础,以机理模型认识和专家先验知识为约束,构建压裂仿真环境,训练泵注决策智能体,可以实现泵注参数实时优化决策与地面装备闭环调控;(3)压裂泵注智能决策系统主要通过压裂工况自主判识、裂缝扩展动态感知与风险预警、压裂泵注参数优化决策、地面装备自主调控、压裂泵注数字孪生等技术实现系统功能。结论认为,基于强化学习理论的压裂泵注智能决策设计理念,有望为加快压裂泵注智能化进程提供有益参考,有助于油气的高效开发。The hydraulic fracturing pumping in the process of reservoir stimulation is characterized by complex and diverse technologies,irreversible process and high risk,so it needs timely and accurate decision-making and control to enhance the effectiveness of the fracturing stimulation.The current decision-making approach mainly takes the hydraulic fracturing design and pumping program as the base to manually control pumping parameters.The advancement of artificial intelligence technology enables the construction of intelligent decision-making systems for hydraulic fracturing pumping and the implementation of autonomous decision-making.Based on a review of researches on hydraulic fracturing pumping decision-making around the world,this paper proposes an intelligent decision-making design concept for hydraulic fracturing pumping based on reinforcement learning theory,and discusses its realization paths.The following results are obtained.First,the core of the intelligent decision-making system for hydraulic fracturing pumping is the pumping decision-making agent,which autonomously determines the pumping program and optimizes the parameters,while transmitting the decision-making actions to surface equipment to form a closed-loop control system,thereby achieving"auto driving"throughout the entire pumping process.Second,the design concept of the intelligent decision-making system for hydraulic fracturing pumping based on reinforcement learning theory is to construct a fracturing simulation environment with the reinforcement learning as the core,the hydraulic fracturing big data as the base,and the mechanism model insights and expert prior knowledge as the constraints,to train the pumping decision-making agent,so as to achieve the real-time optimization of decision and closed-loop feedback control of ground equipment according to pumping parameters.Third,the intelligent decision-making system for hydraulic fracturing pumping achieves its functionality primarily through the technologies such as autonomous identification of fra

关 键 词：水力压裂 泵注 人工智能 决策优化 决策系统 关键技术 路径 

分 类 号：TE377[石油与天然气工程—油气田开发工程]