基于IBES-TPGM(1,1,λ,η)模型的原油管道结蜡厚度预测研究  

Prediction of wax deposition thickness in crude oil pipelines based on an IBES−TPGM(1,1,λ,η)model

作　　者：赵梦龙 程长坤 李本全[1] 沈慧 李春保 宋恺 ZHAO MengLong;CHENG ChangKun;LI BenQuan;SHEN Hui;LI ChunBao;SONG Kai(Oil&Gas Technology Research Institute;Youth League Committee Office,PetroChina Qinghai Oilfield Company,Jiuquan 736202,China)

机构地区：[1]中国石油青海油田公司油气工艺研究院,酒泉736202 [2]中国石油青海油田公司团委办公室,酒泉736202

出　　处：《北京化工大学学报(自然科学版)》2025年第2期15-25,共11页Journal of Beijing University of Chemical Technology(Natural Science Edition)

基　　金：中国石油科技创新基金研究项目(2023-kj-13D)。

摘　　要：为了准确预测原油管道结蜡厚度的变化趋势,解决传统灰色模型(GM)在模型结构和数据处理方面上的不足,构建了三参数灰色模型(TPGM)(1,1),从初始化种群和增加莱维飞行策略等方面对秃鹰搜索算法(BES)进行改进,利用改进的秃鹰搜索算法(IBES)对TPGM(1,1)模型的初始值与背景值进行自动寻优,建立了基于IBES-TPGM(1,1,λ,η)的管道结蜡厚度优化预测模型。利用室内环道和现场管道的结蜡厚度数据进行训练和预测,对比了不同模型的预测精度,并分析了数据集数量对模型精度的影响。分策略消融实验表明,IBES算法在寻优精度、收敛速度和全局搜索能力上均优于其他算法,改进策略具有实用性;IBES-TPGM(1,1,λ,η)模型在室内环道实验数据和现场管道数据上的平均相对误差最小,分别为0.6867%和0.1527%,预测效果优于GM(1,1)、TPGM(1,1)和BES-TPGM(1,1,λ,η)模型。所建模型对于训练集的数量要求较低,适用于管道结蜡厚度的中长期预测,研究结果可为清管周期的确定以及管道的安全运行提供实际参考。In order to accurately predict the variation in wax deposition thickness in crude o il pipelines and allevi-ate the shortcomings of the traditional grey model(GM)in model structure and data processing,a three-parameter grey model(TPGM)(1,1)has been constructed.The bald eagle search algorithm(BES)was improved in terms of initial population and Levy flight strategy.The improved bald eagle search algorithm(IBES)was used to automati-cally optimize the initial value and background value of the TPGM(1,1)model,and an optimization prediction model of pipeline wax deposition thickness based on IBES-TPGM(1,1,λ,η)was established.The wax deposition thickness data for the indoor loop and field pipeline were used for training and prediction.The prediction accura-cies of different models were compared,and the influence of the number of data sets on the accuracy of the model was analyzed.Ablation experiments show that the IBES algorithm is superior to other algorithms in optimization accuracy,convergence speed and global search ability,and the improved strategy is of practical use.The average relative errors of the IBES-TPGM(1,1,λ,η)model on indoor loop experimental data and field pipeline data are the smallest,0.6867%and 0.1527%,respectively.The prediction effect is better than that of GM(1,1),TPGM(1,1)and BES-TPGM(1,1,λ,η)models.The established model has low requirements for the number of training sets and is suitable for medium and long-term prediction of pipeline wax deposition thickness.This work provides a practical reference for the determination of pigging cycles and the safe operation of pipelines.

关 键 词：秃鹰搜索(BES)算法 三参数灰色模型(TPGM)(1 1 λ η) 管道 结蜡厚度 相对误差 

分 类 号：TE832[石油与天然气工程—油气储运工程]