Coriolis Flowmeter for Two-Phase Flow Measurement Based on TCN-LSTM

Coriolis flow meters, are extensively utilized for directly measuring the mass flow of fluids across various industries. However, achieving high-accuracy measurements in two-phase flow remains an unresolved challenge. This study evaluates several commonly used time series models with different signal lengths for input feature variables, aiming to compare their dynamic following capabilities and prediction accuracy. Additionally, a composite model, TCN-LSTM, is introduced to enhance predictive accuracy and tracking ability. The performance of TCN-LSTM is assessed through both offline and online prediction experiments. Online prediction experiments include both static and dynamic scenarios. The online static prediction experiment assesses TCN-LSTM’s prediction ability under relatively stable operating conditions. The online dynamic prediction experiment provides a horizontal comparison and better prediction between the prediction accuracy and follow-up prediction ability of TCN-LSTM and the Stacking model proposed in [4]. The result indicates that TCN-LSTM have a better prediction performance.