Thermodynamically-consistent flash calculation in energy industry: From iterative schemes to a unified thermodynamics-informed neural network

Multicomponent multiphase fluid flows are commonly seen in the engineering practice of hydrocarbon production and transportation; thus, the phase-wise heat and mass transfer mechanisms underneath the macroscopic flow and transport behaviors are essentially needed for better understanding of the physical phenomena and optimization of the industrial processes. Flash calculation, as the main approach computing phase equilibrium conditions, has arisen increasing interests to establish the thermodynamic foundations of multiphase flow simulation, as well as to determine whether two-phase model is needed. In this paper, the general thermodynamically-consistent flash calculation scheme will be developed, and the general adaptability to various special mechanisms will be analyzed. A unified framework of thermodynamics-informed neural network will also be designed to accelerate conventional iterative flash calculation schemes that will be applied in various engineering scenarios to provide certain suggestions to the energy industry based on the predictions and analysis. Novelty Statement A thermodynamically-consistent flash calculation scheme incorporating various special mechanisms that are often met in energy industry. A unified thermodynamics-informed neural network structure for various engineering demands in the energy industry. Suggestions to the energy industry to optimize the productions based on the phase transition predictions and analysis.

Automated summary

This paper presents a thermodynamically-consistent flash calculation scheme for phase transition heat and mass transfer mechanisms in multi-component multiphase fluid flows. It also proposes a thermodynamics-informed neural network framework to accelerate flash calculation. The study provides suggestions for optimizing production in the energy industry.