改进BP神经网络在三相流相分率检测中的应用

本文构建了基于Bayesian正则化算法的BP神经网络模型,对水平管油气水三相流相分率进行了预测,解决了振动管密度法只能测量两相流相分率的问题.通过设计的实验装置获得了测试样本,进行了预测效果检验.结果表明,神经网络预测值与实际值非常吻合,含气率预测最大误差为2.7%,含水率最大误差为3.8%,大大提高了相分率的测量精确度,为利用振动式密度计测量油气水三相流相分率提供了一种有效的方法.

Application of Improved Neural Networks in the Measuring Phase Fraction of Three-phase Flow

To establish neural networks model based on Bayesian regularization algorithm to determine phase fraction of oil-gas-water three-phase flow in horizontal pipe,solving the problem that vibrating tube density approach only measured two-phase flow.The neural networks were trained according to the learning samples obtained from a specially designed device and predicted the results.The results showed that gas fraction was predicted with an error of 2.7% and water fraction with an error of 3.8%.The measuring accuracy of phase fraction was improved greatly.This study provided a new way of using vibrating densitometer to measure the phase fraction of three-phase flow.