Simulating hydraulics of large water drainage systems in HPC environments using the new SWMM5+ engine

Edward Tiernan, Cheng-Wei (Justin) Yu and Ben Hodges


Drainage network simulation tools need to transition to parallel programming techniques in order to take advantage of computational trendstowards multi-/many-processor computing environments. The SWMM5+ hydraulics computational engine is a public-domain, parallelized hydraulic simulator that uses a no-neighbor discretization and finite-volume formulation of the Saint-Venant equations. SWMM5+ acts as an independently called module within the larger workflowof EPA SWMM5.The hypothesized value of the SWMM5+ engine is to enable SWMM users to simulate large systems at speeds approaching real-time simulation.

A proof of concept study was developed to test the limitations of the new SWMM5+ engine on large-scale systems. Weused continuous simulation (6 months) ofcoastal river basinsin Texas as the test case. By approximating this systemwith trapezoidal cross-sections, we sought to avoidadditional hydraulic complexity contributed byhigh-resolution cross-sections or engineered urban drainage practices. National Water Model forcing data was used to simulate real-world hydrologiccomplexity. The hydraulic performance of the SWMM5+ engine on this test-case was compared with an established model, the Simulation Program for River NeTworks (SPRNT); simulations of the system using EPA SWMM5 kinematic and dynamic wave settings were also included in the comparison. The SWMM5+ simulations were conducted in the high-performance computing (HPC) environment of the Stampede2 super-computer at the Texas Advanced Computing Center (TACC).

The results show that SWMM5+ can achieve reasonable agreement with SPRNTand the EPA SWMM5 kinematic wave formulation. The EPA SWMM5 dynamic wave formulation, finicky as always, did not produce stable results for mostcompletedsimulations. Preliminary observations suggest that differences in the numerical architecture of each simulation program become exacerbated when downstream portions of the system are observed.

 Click here to watch recorded presentation on YouTube.

Permanent link: