基于优化BP神经网络的输油管道原油温降计算研究  

Study on Calculation of Crude Oil Temperature Drop in Oil Pipeline Based on Optimized BP Neural Network

作　　者：杨林 YANG Lin(Daqing Oilfield Design Institute Co.,Ltd.)

机构地区：[1]大庆油田设计院有限公司,黑龙江省大庆市163712

出　　处：《油气田地面工程》2023年第3期35-39,61,共6页Oil-Gas Field Surface Engineering

摘　　要：热油管道温降计算是制定合理的设计、生产运行方案和节能降耗的关键技术环节。传统的温降计算模型计算结果存在着较大偏差。为提高计算精度,利用输油管道数字化普及所积累的大量原始生产运行数据,采用机器学习方法,开展输油管道温降计算模型研究。以某热油管道为研究对象,建立了传统温降拟合计算模型和优化BP温降计算模型,并采用自适应权值函数和自适应变化激活函数对BP神经网络进行了改进。分别采用两种模型对管道温降进行计算,并与实测数据进行对比,结果表明:优化BP温降计算模型与传统拟合温降计算模型相比,均方根误差由0.0511降为0.0384,决定系数R2由0.8827降为0.9746,平均相对误差降低了10.41%,计算精度显著提高,有效提高了输油管道温降预测的准确性。The temperature drop calculation of hot oil pipelines is a key technical link in formulating reasonable design,production and operation plans,and energy conservation and consumption reduction.The calculation results of traditional temperature drop calculation models have significant deviations.In order to improve the calculation accuracy,a large number of original production and operation data accumulated by the popularization of oil pipeline digitalization are used,and machine learning methods were to carry out the research on the calculation model of oil pipeline temperature drop.Taking a hot oil pipeline as the research object,the traditional temperature drop fitting calculation model and the optimized BP temperature drop calculation model are established,and the BP neural network is optimized by using the adaptive weight function and the adaptive change activation function.Two models are used for calculation,and compared with the measured data.The results show that,compared with the traditional fitting temperature drop calculation model,the mean square error of the optimized BP temperature drop calculation model is reduced from 0.0511 to 0.0384,the determination coefficient R2 is reduced from 0.8827 to 0.9746,and the average relative error is reduced by 10.41%.The calculation accuracy is significantly improved,which effectively improves the accuracy of pipeline temperature drop prediction.

关 键 词：热油管道 温降 拟合计算模型 神经网络 BP模型 

分 类 号：TE832[石油与天然气工程—油气储运工程] TP183[自动化与计算机技术—控制理论与控制工程]