BP神经网络模型在管道腐蚀风险智能预测中的应用  

Application of BP Neural Network Model in Intelligent Prediction of Pipeline Corrosion Risk

作　　者：肖雯雯[1] 李俊[1] 梁婷婷 高思哲 胡艳萍 王震[2] XIAO Wenwen;LI Jun;LIANG Tingting;GAO Sizhe;HU Yanping;WANG Zhen(SINOPEC Northwest Oilfield Company,Urumqi,Xinjiang 830011,China;SINOPEC Jianghan Petroleum Engineering Design Company Limited,Wuhan,Hubei 430073,China)

机构地区：[1]中国石化西北油田分公司,新疆乌鲁木齐830011 [2]中国石化江汉石油工程设计有限公司,湖北武汉430073

出　　处：《石油管材与仪器》2021年第6期56-61,共6页Petroleum Tubular Goods & Instruments

摘　　要：通过对5种常见管道腐蚀风险预测方法的适用范围、可靠性和经济性进行比选,确定出了适用于塔河油田的腐蚀风险预测方法。借助遗传算法对经典BP神经网络算法进行了优化,有效提高了BP神经网络的准确性和可靠性。建立了基于改进神经网络模型的管道腐蚀速率计算方法,根据腐蚀速率对管道腐蚀程度进行风险分级,结合塔河油田管道类型、H_(2)S含量、人口密集度、自然环境获得风险发生的严重性分级,最终获得管道风险度和风险级别。识别出目前管道腐蚀穿孔泄漏安全风险在12区、环保风险在塔河油田TP区块。分别选取7个区块中的典型管线开展腐蚀风险评价,21条管道中有8条管道风险等级处于高风险,13条管道风险等级为中风险。采用非开挖磁力层析组合检测技术对6处高腐蚀穿孔风险点进行分析,验证了方法的准确性。By comparing the applicable scope, reliability and economy of five common pipeline corrosion risk prediction methods, a corrosion risk prediction method suitable for Tahe Oilfield was determined. The classical BP neural network algorithm is optimized with the help of genetic algorithm, which effectively improves the accuracy and reliability. A set of improved neural network model was introduced to establish the pipeline corrosion rate method. Corrosion risk classification was obtained according to the corrosion rate, the severity of the risk occurrence classification was obtained considering pipeline type, H_(2)S content, population density and natural environment in Tahe Oilfield, and finally the pipeline risk degree and level were determined. It was identified that the current pipeline corrosion perforation leakage risk is in zone 12 and the environmental risk is in the TP zone. The typical pipelines in 7 blocks were selected for corrosion risk assessment respectively, which showed that 8 of the 21 pipelines were at high risk level and 13 pipelines were at medium risk. The accuracy of this method was verified by analyzing 6 high corrosion perforation risk points using non-dig magnetic tomography detection technology.

关 键 词：神经网络 集输管道 风险分析 智能预测 

分 类 号：TE988[石油与天然气工程—石油机械设备]