稠油开采注汽锅炉在线综合预警关键技术  

Research on key technology for online comprehensive early warning of steam-injection boilers in heavy oil exploitation

作　　者：王军茹[1] 吴昊洋 王军平 易军凯 WANG Junru;WU Haoyang;WANG Junping;YI Junkai(School of Automation,Beijing Information Science&Technology University,Beijing 100192,China;Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]北京信息科技大学自动化学院,北京100192 [2]中国科学院自动化研究所,北京100190

出　　处：《哈尔滨工程大学学报》2024年第11期2218-2225,共8页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(61573230);北京高等学校高水平人才交叉培养“实培计划”(科研)项目(202011232303);教育部产学合作协同育人项目(201901009004).

摘　　要：为了对工作在高温高压下注汽锅炉的各项运行参数进行在线准确监测和异常预警,本文在对稠油开采注汽锅炉工况参数进行采集、处理、分析的基础上,提出对注汽锅炉显性故障和隐性故障进行检测的方案。采用长期短期记忆神经网络,利用锅炉的时序数据对系统进行分析和建模,完成锅炉显性故障检测和预警,并通过预测数据的方式缓解锅炉大时滞的特性;利用深度异常检测技术,将无故障判别标准的数据进行隐性故障分析和预警。本文提出的综合预警方案对克拉玛依油田注汽锅炉进行了实验验证,预测误差仅有0.08%,同时异常检测范围也在设定值范围内。This study aims to accurately monitor and provide abnormal warnings for various operating parameters of steam-injection boilers working at high temperatures and pressures.The operating parameters of steam-injection boilers used for heavy oil extraction are collected,processed,and analyzed,and according to the obtained data,a scheme for detecting explicit and implicit faults in these boilers is proposed.First,the LSTM neural network is used to analyze and model the system using the boiler time series data.This enables explicit fault detection and early warning while also alleviating the issue of large boiler time delays through predictive data analysis.Second,deep anomaly detection technology is employed to perform implicit fault analysis and early warning based on data from fault-free discrimination standards.The proposed comprehensive early warning scheme is validated through experiments on steam-injection boilers in the Karamay Oilfield.The prediction error is only 0.08%,and the anomaly detection range is within the set value limits.

关 键 词：稠油开采 注汽锅炉 大时滞 神经网络 时序数据 显性故障 隐性故障 在线监测 异常预警 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]