石油化工高等学校学报2017，Vol.30Issue(5)：80-85,6.DOI:10.3969/j.issn.1006-396X.2017.05.015
基于支持向量机-CV的天然气水合物生成预测
Prediction of Natural Gas Hydrate Formation Based on Support Vector Machine (SVM)-CV
摘要
Abstract
Natural gas hydrate has the advantages of abundant reserves,large calorific value and low emission,which can mitigate the environmental pollution problems caused by traditional fossil energy.The generation process of natural gas hydrate form is a system with multi-components and multi-physical states.The nucleation process is complex,which needs to consider the effects of pressure,temperature,promoters,stirring speed and so on.It is difficult to accurately predict the hydrate formation,because the hydrate formation process not only involves thermodynamics problems but also dynamics problems.In our paper,the support vector machine theory combined with experimental data was used to establish support vector machine prediction model for predicting natural gas hydrate equilibrium pressure.The prediction accuracy was estimated by using the mean square error,the square correlation coefficient,the square absolute percentage error and the average absolute error.The results are 8.37008×10 -5 ,99.8976%,0.5424%,1.9900%,respectively.The pre-treatment origin data were normalized ([1, 2])and the nuclear parameter g (4 )and punishment factor c (1.4142 )were optimized by using cross validation methods. Simulation results show that the equilibrium pressure obtained by support vector prediction model is good in agreement with the equilibrium obtained by experiments.The better ideal prediction effects prove that the model has advantages of accuracy and reliability.It can provide certain reference for research on natural gas hydrate in future.关键词
天然气水合物/结晶/成核/热力学/支持向量机/交叉验证Key words
Natural gas hydrate/Nucleation/Dynamics/Thermodynamics/Support vector machine/Cross validation分类
能源科技引用本文复制引用
宫清君,马贵阳,潘振,刘培胜,李存磊..基于支持向量机-CV的天然气水合物生成预测[J].石油化工高等学校学报,2017,30(5):80-85,6.基金项目
国家自然科学基金项目(41502100) （41502100）
辽宁省高等学校优秀人才支持计划项目(LJQ2014038). （LJQ2014038）
