基于相关向量机算法的埋地热油管道温降预测

Prediction for Temperature Drop of Buried Hot-oil Pipeline Based on the Relevant Vector Machine

温降是热油管道运行管理中优化输送方案及分析运行能耗的决定因素,针对传热系数难以获取及运行参数波动引起的无法对温降进行准确预测的问题,对埋地热油管道温降与多种运行参数及油品物性之间的相关性进行分析,提出一种基于相关向量机算法(RVM)的埋地热油管道温降预测的新方法.通过对出站油温、出站压力、输量、地表温度、埋深、管长、管径和油品物性与温降之间的内在规律进行学习训练相关向量机,建立埋地热油管道温降预测的相关向量机模型.对东北某输油管道温降进行预测的结果表明,方法与传统的反算插值法相比,预测结果平均相对误差降低4.43%,具有预测精度高、泛化性好等优点,更适用于现场实际工况下的埋地热油管道温降的预测.

Temperature drop is the decisive factor for optimizing transporting schemes and analyzing consumptions in the management of hot-oil pipelines.To solve difficulties to get coefficient of heat transfer and to predict the temperature drop precisely,a new method is put forward,i.e.temperature drop prediction for buried hot-oil pipeline based on the Relevant Vector Machine.According to the inherent laws learning of Relevant Vector Machine between the relations of temperature drop and outgoing oil temperature,pressure,output,surface temperature,buried depth,pipe length,pipe diameter and oil property.The results come after the survey for an oil pipeline temperature drop in Northeast China,deviation decreased by 4.43%using the new methods the author set than traditional inverse calculation difference.Besides,the method features high prediction accuracy,high generalization,etc,which is more favorable when it＇s used in real temperature drop predictions for buried hot-oil pipeline.