Parallel computation of dynamic simulation of cool down process in LNG cold box

Key words:  natural gas liquefaction  cool down process in cold box  dynamic simulation  multiple instruction multiple data(MIMD)  multi-core  parallel computation

Abstract:

As the single-core can 't afford the high load of dynamic simulation of the cool down process in LNG cold box, a symmetric multi-core parallel computing method under shared memory was proposed to implement the dynamic simulation of the cool down process in cold box of nitrogen expansion liquefaction process with pre-cooling. In dynamic mathematical models, one-dimensional model is used in the simulation of plate-fin heat exchanger, and compressor, expander and throttle were treated as steady-state operations. In parallel methods, compute units were divided and coupled in the boundary based on the characteristics of parallel machine and liquefaction process. The simulation processes of parallel units were controlled using synchronization barrier, and the data were communicated through explicit function calls. The comparison of simulation results with experimental data shows that the dynamic models and parallel methods were reasonable. The simulation results show that the cooling rate of the working fluid depends mainly on the polytropic efficiency of the expanders. The result of performance test of parallel computation shows that the parallel method accelerates the simulation process 23 times, and the efficiency of computing nodes is 3.83 times the efficiency of the single-core computing.