Abstract / Description of output
Accurate multiphase flow measurement is vital in monitoring and optimizing various production processes. Deep learning has as of late arose as a promising approach for assessing multiphase flowrate dependent on various customary flow meters. In this paper, we propose a multi-modal sensor and Temporal Convolution Network (TCN) based method to predict the volumetric flowrate of oil/gas two-phase flows. The volumetric flowrates of the liquid and gas phase vary from 0.96 - 6.13 m3/h and 5.5 - 121.2 m3/h, respectively. The multi-modal sequential sensing data are simultaneously collected from a Venturi tube and a dual-plane Electrical Capacitance Tomography (ECT) sensor in a pilot-scale multiphase phase flow facility. The reference data are derived from the single-phase flowmeters. Z-score and First-Difference (FD) data pre-processing methods are employed to manipulate the collected instantaneous time series multi-modal sensing data. The pre-processed data are utilized for training the TCN model. Experimental results reveal that the TCN model can effectively predict the multiphase flowrate based on the multi-modal sensing data.The results provide guidance on data pre-processing methods for multiphase flowrate estimation and demonstrate the effectiveness of combining multi-modal sensors and TCN for multiphase flowrate prediction under complex flow conditions.
Original language | English |
---|---|
Journal | IEEE Sensors Journal |
Early online date | 29 Apr 2022 |
DOIs | |
Publication status | E-pub ahead of print - 29 Apr 2022 |
Keywords / Materials (for Non-textual outputs)
- Temporal Convolutional Network (TCN)
- Z-Score
- first difference method
- multiphase flowrate measurement
- time series data