当前位置：首页 > 过刊浏览->2025年第36卷第6期
基于机器学习的H2S/CO2耦合腐蚀速率预测
Prediction of H2S/CO2 Coupling Corrosion Rate Based on Machine Learning
浏览(7) 下载(3)
- DOI:
- 作者:
- 李嘉轩1，孙庆峰2?，刘瀚宇3，刘怀珠4，王占荣5，王 强5，陈良超1
Li Jiaxuan 1, Sun Qingfeng 2?, Liu Hanyu 3, Liu Huaizhu 4, Wang Zhanrong 5, Wang Qiang 5,Chen Liangc
- 作者单位:
- 1.北京化工大学 机电工程学院；2.山东港源管道物流有限公司；3.中国特种设备检测研究院；4.唐山冀油瑞丰化工有限公司；5.榆林市特种设备检验检测院
1. School of Mechanical and Electrical Engineering, Beijing University of Chemical Technology; 2. Shandong Gangyuan Pipeline Logistics Co., Ltd. 3. China Special Equipment Inspection and Research Institute; 4. Tangshan Jiyou Ruifeng Chemical Co., Ltd.; 5. Yulin Special Equipment Inspection and Testing Institute
- 关键词:
- H2S/CO2腐蚀；腐蚀机理；腐蚀速率预测；机器学习
H2S/CO2 corrosion; corrosion mechanism; corrosion rate prediction; machine learning
- 摘要:
- 摘要：中国大多数油田进入中后期开发阶段，由于油井长时间处于含腐蚀性介质环境下，油气输送管道频繁发生腐蚀、开裂以及泄漏等问题。针对油气输送管道腐蚀等问题，明确腐蚀机理及影响因素，收集相关运行数据，筛选影响或表征腐蚀的主要因素，建立基于 BP神经网络等机器学习算法的腐蚀速率预测模型。结果表明，设备中含有H2S/CO2等腐蚀性介质是发生腐蚀的主要原因，经训练得到的BP神经网络模型的决定系数为0.847，均方误差为2.930，平均绝对误差为1.391，预测准确度优于其它2个模型，可为解决石油化工装置的腐蚀问题及研究提供参考。
Abstract: Most oil fields in China have entered the mid-to-late development phase. Due to the long-term exposure of oil wells tocorrosive media, oil and gas pipelines frequently experience corrosion, cracking, and leakage. To address these issues, this studyclarified the corrosion mechanisms and influencing factors, collected relevant operational data, screened the key factorsinfluencing or characterizing corrosion, and established a corrosion rate prediction model based on machine learning algorithms,such as BP neural networks. Results indicate that the presence of corrosive media, such as H2S / CO2, in the equipment is theprimary cause of corrosion. The trained BP neural network model achieved a coefficient of determination of 0.847, a mean squareerror of 2.93, and a mean absolute error of 1.391, outperforming the other two models in prediction accuracy. This study provides areference for addressing corrosion issues and research in petrochemical plants.