摘要
通过多传感器技术对原油含水率测量有影响的多个参量进行测定,提出基于多元非线性回归和神经网络融合处理两种方法建立原油含水率预测模型,并采用分段建模的方法分别进行改进。评价结果表明:神经网络模型预测效果优于多元非线性回归模型,原油含水率分段预测模型效果优于统一模型。尤其是改进的神经网络分段预测模型具有网络结构简化、收敛速度快,泛化能力强的特点,取得很好的拟合精度和预测效果。
Using multi-sensor technology, some parameters affecting the measurement of water content of crude oil are measured, and prediction models of water content of crude oil based on the methods of multivariate nonlinear regression and artificial neutral networks are presented, and then being improved by subsection modeling. The assessed resuits show that the prediction effect of artificial neutral networks model is better than that of multivariate nonlinear regression model, as well as the forecast effect of subsection model for water content of crude oil is better than that of united model. The improving neural network subsection prediction model can take advantage of simple network structure,fast convergence rate and strong generalization capability, and get a good modeling effect.
出处
《化工自动化及仪表》
EI
CAS
2006年第4期61-63,70,共4页
Control and Instruments in Chemical Industry
关键词
原油
含水率
多元非线性回归
神经网络
预测模型
crude oil
water content
multivariate nonlinear regression
neutral networks
prediction model