摘要
为了提高输油管道运行电耗的预测准确度,给生产调度和节能开泵方案提供实际指导。以某油田两条输油管道为例,在收集原始数据的基础上,根据水力学原理对变量进行扩充,得到更多与管道电耗相关的特征,再利用K近邻互信息估计对变量进行特征选择,最后通过支持向量机(SVM)实现小样本信息的回归预测。结果表明,影响管道电耗的因素依次为日输量、雷诺数、进出站压差、出站压力、进站压力等,特征选择结果较皮尔逊相关系数法相比,更为合理、可信;对比了多种模型的预测效果,管道A和管道B分布在5#模型和4#模型上的相对误差范围最小,模型的泛化能力最强;管道A每月可节约电耗0.26×10^(4)~0.67×10^(4)kWh,管道B每月可节约电耗0.14×10^(4)kWh。
In order to improve the prediction accuracy of power consumption of oil pipeline opera-tion and provide practical guidance for production scheduling and energy conservation start-up scheme,taking two oil pipelines in an oilfield as an example,on the basis of collecting original data,the variables are expanded according to hydraulics principal to get more features related to pipeline pow-er consumption,and then the characteristics of variables are selected by using K-nearest neighbor mu-tual information estimation.Finally,the regression prediction of small sample information is realized by support vector machine(SVM).The results show that the factors affecting the power consumption of pipeline are daily transmission,Reynolds number,inlet and outlet differential pressure,outlet pres-sure,inlet pressure and so on.Compared with the Pearson correlation coefficient method,the selec-tion results are more reasonable and reliable.After comparing the prediction results of various models,#the relative error range of Pipeline A and Pipeline B is the smallest on Model 5 and Model 4#,and the generalization ability of the model is the strongest.Even more to the point,the monthly electricity 44 consumption of Pipeline A can be saved 0.26×10^(4) kWh~0.67×10^(4) kWh while the monthly electricity 4 consumption of Pipeline B can be saved 0.14×10^(4) kWh.
作者
林炜国
LIN Weiguo(Petrochina Electric Energy Infrastructure Management Center of Daqing Oilfield Co.,Ltd.)
出处
《石油石化节能与计量》
2024年第1期32-37,共6页
Energy Conservation and Measurement in Petroleum & Petrochemical Industry
关键词
输油管道
电耗
预测
互信息估计
支持向量机
oil pipeline
power consumption
prediction
mutual information estimation
support vector machine
