改进天鹰优化器优化Elman神经网络模型预测管道蜡沉积速率  

Prediction of Wax Deposition Rate in Pipeline by Optimizing Elman Neural Network Model with Improved Aquila Optimizer

作　　者：陈卓 刘波 张源 杨云博 田洋阳[5] CHEN Zhuo;LIU Bo;ZHANG Yuan;YANG Yun-bo;TIAN Yang-yang(Sinopec Northwest Oilfield Company,No.2 Oil Production Plant,Urumqi 830016,China;The Changqing Sub-company of Northern Pipeline of State Pipeline Incorporation,Yinchuan 750001,China;PetroChina Changqing Oilfield Company,No.6 Gas Production Plant,Xi'an 710021,China;PetroChina Changqing Oilfield Company,Changbei Operation Company,Xi'an 710018,China;College of Petroleum Engineering,Xi'an Shiyou University,Xi'an 710065,China)

机构地区：[1]中国石化西北油田分公司采油二厂,乌鲁木齐830016 [2]国家管网北方管道长庆输油气分公司,银川750001 [3]中国石油长庆油田分公司第六采气厂,西安710021 [4]中国石油长庆油田长北作业分公司,西安710018 [5]西安石油大学石油工程学院,西安710065

出　　处：《科学技术与工程》2023年第19期8179-8186,共8页Science Technology and Engineering

基　　金：陕西省自然科学基础研究基金(2021JQ-602);陕西省教育厅科研计划(21JK0831)。

摘　　要：管道结蜡一直是困扰含蜡原油输送的棘手问题,建立准确的蜡沉积速率预测模型对于保障管道的安全运行具有重要的实际意义。考虑到Elman神经网络(Elman neural network,ENN)模型的不足(易陷入极小点、泛化能力弱),基于蜡沉积速率的主要影响因素,提出了一种基于改进天鹰优化器(引入鲸鱼优化算法的狩猎策略对天鹰优化器的局部搜索能力进行改进)的ENN模型,并基于两组室内实验数据对比分析了所建新模型和其他模型预测精度的差异。结果表明,改进新模型的平均绝对百分比误差分别为0.7603%、1.2452%,其预测精度明显高于传统ENN模型;采用改进天鹰优化器建立的ENN模型可对初始权值和阈值进行寻优处理,极大提高了泛化能力,因此具有预测精度高的优点。Pipeline wax deposition has always been a thorny problem perplexing the transportation of waxy crude oil.It is of great practical significance to establish an accurate wax deposition rate prediction model for ensuring the safe operation of the pipeline.Considering the shortcomings of Elman neural network(ENN)model(easy to fall into minimum points and weak generalization ability).Based on the main influencing factors of wax deposition rate,an ENN model based on the improved Aquila optimizer(the hunting strategy with whale optimization algorithm improves the local search ability of Aquila optimizer)was proposed.Based on two groups of indoor experimental data,the difference of prediction accuracy between the new model and other models was compared and analyzed.The results show that the mean absolute percentage error of the improved new model is 0.7603%and 1.2452%respectively,and its prediction accuracy is obviously higher than that of the traditional ENN model.The ENN model established by the improved Aquila optimizer can optimize the initial weights and thresholds,which greatly improves the generalization ability,so it has the advantage of high prediction accuracy.

关 键 词：含蜡原油 蜡沉积速率 改进天鹰优化器 ELMAN神经网络 预测 

分 类 号：TE927[石油与天然气工程—石油机械设备]