Machine learning for multiphase flowrate estimation with time series sensing data
Haokun Wang, Maomao Zhang, Yunjie Yang
Abstract
In this paper, we investigate the prediction of multiphase flowrate based on multi-modal time series sensing data by using machine learning. The time series differential pressure data generated from Venturi tube, and pressure, temperature data are employed as network input. We implement and compare the performance of three machine learning methods including Deep Neural Network (DNN), Support Vector Machine (SVM) and Convolutional Neural Network (CNN). The multi-modal multi-phase flow sensing data are collected in a laboratory-scale flow facility under various flow conditions. Moving average of the collected instantaneous sensing data is applied to train the developed DNN, SVM and CNN. The result analysis shows that DNN and SVM methods can achieve satisfactory liquid and gas flowrate prediction accuracy under various flow conditions, such as different water in liquid ratio, different gas volume fraction and different flow regimes.