基于LCASO-BPNN模型的单质硫溶解度预测  

Prediction of elemental sulfur solubility based on LCASO-BPNN model

作　　者：汪洋[1] 陈俊杰[1] 谢梦雨 何巾国 赵浩童 贺三[2] 申小冬 WANG Yang;CHEN Junjie;XIE Mengyu;HE Jinguo;ZHAO Haotong;HE San;SHEN Xiaodong(Northeast Sichuan Gas Field,PetroChina Southwest Oil&Gas Field Company,Dazhou 635000,Sichuan Province,China;School of Petroleum and Natural Gas Engineering,Southwest Petroleum University,Chengdu 610500,Sichuan Province,China;Offshore Oil Engineering Co.,Ltd.,Tianjin 300450,China;College of Energy,Chengdu University of Technology,Chengdu 610059,Sichuan Province,China)

机构地区：[1]中国石油西南油气田公司川东北气矿,四川达州635000 [2]西南石油大学石油与天然气工程学院,四川成都610500 [3]海洋石油工程股份有限公司,天津300450 [4]成都理工大学能源学院,四川成都610059

出　　处：《化学工程》2023年第12期56-61,共6页Chemical Engineering(China)

基　　金：四川省科技厅应用基础研究项目(2020YJ0393)。

摘　　要：使用智能算法对硫溶解度进行预测是分析解决硫沉积问题的重要路径之一。为提高算法精度,提出一种采用基于混沌理论与Logistic映射改进的原子搜索优化算法对BP神经网络的权值和阈值进行优化的LCASO-BPNN预测模型,考虑温度、压力及CH_(4)、H_(2)S、CO_(2)摩尔分数5个影响硫溶解度的因素,选用224组实验数据对模型进行训练与预测,使用EAARD(平均绝对相对偏差)、ERMSE(均方根误差)、ESD(标准偏差)和测定系数R^(2)这4个评估参数对模型进行评估。模拟结果表明:提出的LCASO-BPNN预测模型的EAARD为4.60%,ERMSE为0.0367,ESD为0.0689,R^(2)为0.9978。较之前的研究,LCASO-BPNN模型具有预测精度高、误差小、模型简便的优势,可应用于实际工程。Predicting sulfur solubilitys by intelligent algorithms is one of the ways to solve sulfur deposition problems.To improve the accuracy,a LCASO-BPNN prediction model was proposed which used an improved atomic search optimization algorithm based on chaos theory and Logistic mapping to optimize the weights and thresholds of BP neural networks.Five factors(temperature,pressure,CH_(4),H_(2)S and CO_(2)molar fracrtion)were considered and 224 sets of experimental data were selected to train and predict the model.E AARD(average absolute relative deviation),E RMSE(root mean square error),E SD(standard deviation)and measurement coefficient R^(2)were used to evaluate the model.The results show that E AARD of the LCASO-BPNN prediction model is 4.60%,E RMSE is 0.0367,E SD is 0.0689,and R^(2)is 0.9978.Compared with previous studies,the LCASO-BPNN model has higher prediction accuracy,smaller error,and is simpler to be applied in practical projects.

关 键 词：硫溶解度 BP神经网络 混沌理论 LOGISTIC映射 原子搜索优化算法 

分 类 号：TE832[石油与天然气工程—油气储运工程]